bp - multiphase design manual

Section 1. Multiphase Fundamentals


Introduction to Multiphase Flow

1.1 Multiphase Flow Nomenclature and Definitions

1.2 Flow Regimes

Issue 2 - December 1994

Section 1. Multiphase Fundamentals BP Multiphase Design Manual

THIS PAGE ISINTENTIONALLY

LEFT BLANK

-

2 Issue 2 - December 1994

BP Multiphase Design ManualI

Section I. Multiphase Fundamentals

1.1 Multiphase Flow Nomenchture and Definitions

Multiphase flow employs a wide range of concepts and nomenclature that require explanationfor the engineer new to the area.

1.1.1 Volume formation factor and oil specific gravity

In a multiphase line gas dissolves in the oil according to the local pressure and temperature.The solubility of gas in the oil increases with increasing pressure and decreasing temperature.Hence the density of the oil, and the volume it occupies, is dependent on the local pressureand temperature. The reference condition at which the oil density is specified is:

Pressure = 14.7 psiaTemperature = 60°F

This is known as stock tank conditions (STC). The oil specific gravity is density of oil at stocktank conditions divided by the density of pure water at stock tank conditions.

Density oil at STCOil SG =

Density water at STC

An oil production rate of 1 stock tank barrel per day (1STBD) is that flowrate which whenproduced to a stock tank at 14.7 psia and 60°F yields 1 barrel (1 bbl i.e. 5.614 ft3) of oil.Hence, the flowrate in STBD is a measure of the stabilised (degassed oil) that would be yieldedif the pipeline flow discharged to a stock tank.

The ratio of volume occupied by liquid at pipeline conditions to that occupied at stock tankconditions is the volume formation factor, B,:

Pipeline barrels (pbbl)B, =

Stock tank barrels (STB)

Hence if the oil flowrate is Q,, (STBD) and the volume formation factor at some pipeline condi-tions is B,, then the actual in-situ volume flowrate of oil Q,, in pbbl/day is given by:

QOA = Q,, x B,

The value of B, usually lies in the range 1 .O to 2.0. A more detailed discussion of the volumeformation factor is given in the Physical Properties section of this manual, Section 2.3.

1.1.2 Water cut and water volume formation factor

The water cut is normally defined as the water fraction of the total liquid flow expressed as apercentage:

Issue 2 - December 1994 3

ISection 1. Multiphase Fundamentals BP Multiphase Design Manual

Qw.3Qw.3Water cut = x 100

(Q,,(Q,, + Qws)

where:

q, = water production rate in STBDQ,, = oil production rate in STBD

The water volume formation factor, B,, is the ratio of the volume occupied by water at anypressure and temperature to that occupied at stock tank conditions. Hence the in-situ volumeflowrate of water Q, in pbbl/day is given by:

QWA = Bw x Q,

The value of B, is normally close to 1 .O.

The total actual liquid flowrate in a pipeline, Q,, is simply given by:

QLA = QLA = QoA+QvAQoA+QvA

1.1.3 Producing gas oil ratio and gas specfic gravity

The producing gas oil ratio (GOR) is the total quantity of gas produced when reservoir fluids areflashed to stock tank conditions. The units of GOR are standard cubic feet per stock tank barrel(SCFLSTB); that is, it is the volume of gas produced at 14.7 psia and 60°F associated with theproduction of 1 stock tank barrel of oil.

The specific gravity of the gas is the density of the gas produced at stock tank conditionsdivided by the density of air measured at 14.7 psia and 60°F.

Density of stock tank gasGasSG =

Density of air at 14.7 psia, 60°F

at 14.7 psiaReservoir fluids SO+

Figure 1.

Standard cubic feet gas

Stock tank barrels oil


BP Multiphase Design Manual Section 1. Mu&phase Fundamentals

In practice the GOR of reservoir fluids is dependent on the number of flash stages taken to getdown to stock tank conditions. This is discussed in more detail in Section 2.3.

The solution gas oil ratio of oil (RJ is the amount of gas dissolved in the oil at any pressure andtemperature. Its units are SCF/STB. It represents the amount of gas produced when the pres-sure and temperature of the oil are altered to stock tank conditions. At any temperature andpressure the amount of free gas present in the pipeline is the difference between the values ofGOR and R,. Hence, if the oil flowrate is Q,, (STBD), the flowrate of gas at any temperature andpressure (Q,,) is given by:

Q GS = (GOR - RJ

The units of gas flowrate are standard cubic feet per day, SCFD. To convert this gas flow intoactual cubic feet per day (Q,) at pipeline conditions, an equation of state (EOS) is used. Thesimplest EOS is the engineering EOS:

PV = nzRT-

or

PI VI PA-=-4 T Z2T2

where:

P = pressureV = volumez = compressibilityT = temperaturen = number of moles

State 1 = stock tank conditions, 14.7 psia 60°FState 2 = pipeline conditions

More information is given in Section 2.3 on GOR and Rs. Equations of state are discussedin more detail in Section 2.2.

1.1.4 In-situ gas and liquid density

The density of oil in a pipeline is less than that of stock tank oil due to the presence ofdissolved gas and temperature expansion effects. These factors are taken into account by useof the volume formation factor and solution gas oil ratio. The density of oil at a condition 2 isthen given by:


Sectton 1. Multipbase Fundamentals BP Multiphase Design Manual

where:

P02 = oil density at pipeline conditions (lb/ft3)

POS = oil density at stock tank conditions (lb/ft3)B,, = oil volume formation factor at pipeline conditionsP = gas density at stock tank conditions (lb/f?)Rz = solution gas oil ratio at pipeline conditions (SCFKTB)

If water is present the in-situ liquid density, pc2, is simply a volume weighted average of the oiland water densities.

P02 x QOA + Rwz x %A

PL2 =

QoA + QwA

where:

P = water density at pipeline conditionsQI: and q, are as defined above.

If the simple engineering EOS is used to correct gas density to pipeline conditions, then in-situgas density is calculated from:

p2 To 20

432 = %I0X - X - X -

PO T2 z2

where:

State 0 refers to stock tank conditionsState 2 refers to pipeline conditions

p, = gas densityP = pressureT = absolute temperaturez = compressibility

1.1.5 Superficial liquid and gas velocity

The superficial liquid velocity, VsL, is the velocity the liquid would have if it occupied theentire cross sectional area of the pipe. It is thus defined as:

QLAQLAv,v, = = --

A,A,

where:

Q, = actual liquid flowrate at pipeline conditions (ft3/sec)A, = cross sectional area of pipe (f?)

-


BP Multiphase Design Manual

ISection I. Multipbase Fundamentals

Similarly the superficial gas velocity, V,,, is the velocity the gas would have if it occupied theentire cross section of the pipe:

clGA

v, = -

AP

where:

QGA = actual gas flowrate at pipeline conditions (ft3/sec)

1.1.6 Mixture velocity and no slip hold-up

The mixture velocity in 2-phase flow is simply defined by:

“IT7 = “SL + “SG

where:

V, = mixture velocity (ftkc)

Now as V,, and V,, are given by:

QLA QGA

v -SL = a n d V, = -

A, A,

then:

QLA QGA

v, = v,+v, = -+-

AP AP

So if we define the total actual fluid flowrate CJ, as:

Q, = Q, + QGA

then:

QT

v, = -

A,

Now in any multiphase system the actual velocities of the liquid and gas, V, and V,, will bedependent on their respective volume flowrates and the fraction of the pipe area they occupy.



If the fraction of the pipe occupied by liquid and gas is H, and H,, then the actual liquid and gasvelocities are given by:

QLAQLA vvSLvL=-.-----=-

HLA, HL

QGA VSGv, = ~ = -

HG.A, HG

Now if the liquid and gas velocities are equal, we say there is no slippage between the phases:

v,v, = = v,v,

and so:

VSL VSG VSL VSG- = -or-=p

HL HG HL (1 - HL)

If there is no slippage between the phases the liquid fraction is termed the non-slip liquid holdup, A,. The above equation then becomes:

VSL VSG-=-

XL (1 - AL)

rearranging we obtain:

VSL

A, =

‘SG + ‘SL

thus:

VSL

A, = -

VII7

Hence the non-slip hold-up of liquid is simply the ratio of superficial liquid velocity to mixturevelocity.

-



Section I. Multiphase Fundamentals

1.1.7 Slippage and liquid hold-up

In most 2-phase pipelines the gas travels faster than the liquid:

Under these conditions there is said to be slippage between the phases. As the liquid velocity isless than it would be under non-slip conditions, the liquid hold-up, H,, must increase to a valuein excess of the non-slip value A, in order to maintain continuity:

gas velocity V,gas hold-up H,

When slip occurs: v, ’ VL

and H, > A,

Figure 2.

Note that irrespective of slippage the in-situ liquid flowrate Q, is always given by:

Liquid hold-up H, is an important parameter in 2-phase flow. The liquid hold-up at any condi-tion of flow, pressure and temperature has a major influence on:

l The type of flow that will occur (i.e. flow regime, see Section 1.2)

l Pressure drop

l The slugcatcher volume required at the reception facilities to handle liquid swept outof the line by a pig.

The value of H, is normally determined by use of an empirical correlation such as those ofBeggs & Brill, and Eaton. The correlations typically provide H, as a function of VSL, VSG, pL andp,. More information is given on hold-up correlations in Section 3.1.

1.1.8 Pressure drop

The total pressure drop over a pipeline system, AP, is given by:

BP, = AP, + AP,, + AP,,



where:

AP, = frictional pressure lossAP,, = pressure drop caused by elevational changes

APacc = pressure drop required to accelerate fluidsAP,, is usually insignificant and is not discussed here.

(a) Frictional pressure loss

In single phase flow frictional pressure drop is given by:

fL pupA P , = - ___

d 2

where:

f = friction factor (Darcy-Weisbach)L = length of piped = pipe diameterp = fluid densityu = fluid velocity

Note that the friction factor used throughout this manual is that of Darcy-Weisbach. This is 4times greater than the fanning friction factor, i.e. in laminar flow f is given by:

64f =

Reynolds No.

Use of the Darcy-Weisbach friction factor has the advantage that the term f.Ud represents thenumber of velocity heads lost over any pipe length L.

Losses over fittings are commonly available in the form of velocity heads lost, or loss coeffi-cients, K. Hence friction loss over a fitting is given by:

PU2AP, = K-

2

For homogeneous non-slip 2-phase flow (i.e. where liquid and gas are thoroughly mixed andflow at the same velocity) we could express the friction pressure loss as:

L P”, K2AP, = f,, __

d 2


BP Multiphase Design Manual Section 1. Multipbas e Fundamentals

where:

V, = mixture velocity

P“S = no slip mixture velocity

PL x “SL + P, x “xi

P =“S

“Ill

where;

f”, = friction based on a mixture Reynolds number, Re.

“, x p,.q x dR e =

IJns

where:

P“S = no slip mixture viscosity

where:

uL = in-situ oil viscosity (live oil viscosity)uG = in-situ gas viscosity

For truly homogeneous non-slip 2-phase flow the above equation will give a reasonable esti-mate of AP,. However, in most multiphase systems slippage occurs between the two phases.As a consequence energy is lost as a result of interfacial shear between the phases, anddepending on flow regime, by other mechanisms. The overall effect is that the total frictionalpressure loss is greater than for homogeneous conditions.

The frictional pressure loss in 2-phase flow is normally calculated using an equation of the form:

L P”, “In2A P , = ft, -

d 2

where:

ft, = 2-phase friction factor.

In most 2-phase correlations ftp is calculated using the following equation:

Issue 2 - December 1994 1 1


ftpftp = = mtpmtp f”Sf”S

where:

mtp = an empirically determined Z-phase multiplierf”, = friction factor, normally calculated on basis of homogeneous flow

The various pressure drop correlations use different methods for determining ftp. These arediscussed in more detail in Section 3.1.

(b) Pressure drop due to elevational changes

The elevational pressure drop is given by:

*pe, = ps L g sin 8

where:

L = segment lengthg = acceleration due to gravity0 = angle of segment to horizontalps = in-situ density of 2-phase mixture (i.e. accounting for slip)

PS = PL H, + P,H,

H, = in-situ liquid hold-upH, = in-situ gas hold-up (1 - HL)

Clearly knowledge of liquid hold-up is vital in determining APeL

Section 3.1 of this manual discusses pressure drop and hold-up correlations in more detail andgives recommendations of the best correlations for crude oil-gas systems.


BP Multiphase Design Manuali


1.2 Flow Regimes

1.2.1 Horizontal and Near Horizontal Flow

When liquid and gas flow concurrently in a pipe they can distribute themselves into any one of anumber of flow patterns depending on their respective flow rates, physical properties, pipe sizeand inclination. Figure 3 (below) shows the principal flow patterns, or flow regimes, occurring inhorizontal or slightly inclined pipelines.

SfatifiedWavy Flow

Slug Flow

fl-A n n u l a r F l o w

DispersedBubble Flow

Figure 3. Flow regimes for a horizontal pipe.

These are briefly described below:

(a) Stratified Smooth Flow

At low gas and liquid flowrates the phases segregate with the liquid flowing along the bottom,and the gas flowing through the upper part of the pipe. The interface between the phases issmooth. This type of flow rarely occurs in the field.



(b) Stratlfwd Wavy Flow (Wavy Flow)

As the gas velocity is increased waves form on the liquid surface. The liquid partly climbs thepipe wall to form a crescent.

00Figure 4. Waves on surface. Liquid climbs pipe wall to form a crescent.

This type of flow is common in gas-condensate systems.

The wave height, and level to which the liquid film climbs the surface of the pipe, increase asthe gas velocity is raised. Some liquid is also torn off from the waves to form droplets whichtravel for a distance in the gas core before being deposited back into the liquid flow.

Figure 5. Liquid climbs further up wall. Some liquid entrained in gas core

(Note that in a horizontal line stratified smooth and stratified wavy flow give rise to approxi-mately steady production of gas and liquid. However, where the line undergoes changes ininclination, and where there is a sufficient quantity of liquid present, the liquid accumulation inlow points of the system may produce a blockage causing the flow to stall. The liquid will thenbe intermittently expelled from the dip as a slug. This condition, commonly referred to assurging, is not normally experienced in low liquid loading gas-condensate systems, but canoccur in low velocity crude oil-gas systems such as at Wytch Farm and Welton).

-


-

BP Multiphase Design Manual Section 1. Multiphase Fundamentals

(c) Annular Flow (Annular-Mist)

As the gas velocity is further increased there comes a point where the liquid film forms acomplete annular ring around the surface of the pipe. The liquid film is obviously thickest at thebottom of the pipe. Some liquid is entrained as a mist in the gas core.

0 OQ0 O00 00

00

0 0

QOQ 0 o

Oo Q0

Figure 6. Liquid forms complete annular ring, thickest at bottom. Some liquid entrainedin core

This type of flow can occur in gas-condensate systems at the top end of their throughputrange. The onset of annular flow marks the transition to a condition where the pipe wall is fullywetted. This transition to annular flow thus has implications for corrosion inhibition.

(d) Plug Flow (Elongated Bubble)

Plug flow occurs at low gas and moderate liquid velocities and is characterised by bubbles ofgas distributed in a liquid continuum. Plug flow can occur in low GOR crude oil-gas systems,such as the Wytch Farm Bridport flowlines, and also in moderate to high GOR systems whenoperated at high pressures.

This type of flow does not normally cause any process upsets as the gas plugs are relativelyshort and are produced regularly.

(e) Slug Flow

Slug flow occurs over a wide range of conditions and is hence often encountered in multiphaselines. In slug flow the fluids are ordered as alternate slugs of liquid and bubbles of vapour.

The variation in liquid and gas production rates associated with slug flow can cause upsets atthe process plant. The extent of the upsets will depend to a large extent on the length of theslugs. Note that plug and slug flow are often jointly termed intermittent flow.

(f) Dispersed Bubble Flow

Dispersed bubble flow occurs at high flowrates when the flow induced turbulence causes theliquid and gas to become well mixed. This type of flow is sometimes referred to as froth flow.



I -

) -

I -

1ooc

1oc

ct42g 1(B3u.-s33.o 1‘t$

cz

.l

.Ol

I -

There is some evidence available that suggests that foaming agents can be used to change aslug flow condition to frothy dispersed bubble flow by altering the gas/liquid surface tension.Hence the chemical nature of the fluids can actually have a bearing on the flow regime by inter-acting with the physical forces which tend to separate the liquid and gas.

The transition between each flow regime is usually presented in the form of a two-dimensionalflowmap, see Figure 7 (below). The usual co-ordinates of such a map are superficial liquidvelocity (effectively liquid volume flowrate) and superficial gas velocity.

n 1: strat to Int

+ 2: int to disp bubble

0 3: strat smooth to wavy

A 4: annular flow transition

Dispersed Bubble

AIntermittent

(plug) ( s l u g )Annular Mist

SmoothStratified

.l 1 10 100 1000 1+04

Superficial Gas Velocity (ft/sec)

Figure 7. Taitel-Dukler horizontal flowmap.

The flowmap shown in Figure 7 (above) was produced by the industry standard method ofTaitel-Dukler. BP’s field experience has shown discrepancies between Taitel-Dukler predictionsand measured data. BP Exploration is developing new mechanistic models for determining


rBP Multiphase Design Manual Section 1. Multiphase Fundamentals

flow regimes. Work on the transition marking the onset of wall wetting is now complete and thismodel should now be used in design.

1.2.2 Vertical Flow

in vertical flow gravity acts to produce symmetrical flow about the centre-line so that there is noequivalent of stratified or wavy flow.

Figure 8 (below) shows the principal flow regimes occurring in vertical flow.

Bubble

Flow

tSlugFlow

t t

Chum Annular

Flow Flow

Figure 8. Flow patterns for upward vertical flow.

(a) Bubble Flow

In bubble flow the gas phase is distributed more or less uniformly in the form of discretebubbles in a continuous liquid phase.


sSection 1. Multiphase Fundamentals BP Multiphase Design Manual

(h) Slug Flow

In the slug flow pattern the two fluids redistribute axially so that at any cross-section the flowrates of liquid and gas vary with time. The gas flows largely in a “Taylor bubble”, which occupiesmost of the pipe’s cross-sectional area and can vary in length from the tube diameter to over ahundred diameters. Between the Taylor bubble and the wall, a thin liquid film flows downward.

(c) Churn Flow

Churn flow is somewhat similar to slug flow. It is, however, much more chaotic, frothy, anddisordered. The bullet-shaped Taylor bubble becomes narrow, and its shape is distorted. Thecontinuity of the liquid in the slug between successive Taylor bubbles is repeatedly destroyedby a high local gas concentration in the slug. As this happens and the liquid falls, this liquidaccumulates, forms a bridge, and is again lifted by the gas.

(d) Annular Flow

Annular flow is character&d by continuity in the axial direction of the gas phase in the core.Liquid flows upward, both as a thin film and as droplets dispersed in the gas. Except at thehighest flow rates, the liquid appears to also flow as large fast- moving lumps that are intermit-tent in nature - either large roll waves travelling over the film or a high local concentration ofdroplets.


Section 2. Plysical Properties

Section 2. Physical Properties

Prediction of Gas and Liquid Physical Properties

2.1 Illtroduction

2.2 Compositional Models

2.3 Black Oil Models


I Section 2. Physical Properties BP Multiphase Design Manual


LEFT BLANK

-


BP Multiphase Design Manual Sectton 2. Physcal Properties

2.1 Introduction

Multi-component phase behaviour is a complex phenomenon which requires accurate determi-nation if two-phase pressure loss, hold-up and flow regime are to be determined with anydegree of confidence.

In practice, experimentally determined phase behaviour is often scarce and one has to resort tosome method of prediction. There are two approaches commonly employed in the predictionof hydrocarbon phase behaviour. These are the “black oil” method, which assumes that onlytwo components i.e. gas and liquid, make up the mixture, and the so called “compositional”approach in which each hydrocarbon component is taken into account. The methods havetheir own relative merits and are discussed below.


ISection 2. Physical Properties BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 2. P&y&al Properties

2.2 Compositional Models

Hydrocarbon mixtures display a phase behaviour that is generally represented in the form of aP-T phase diagram, thus:

Figure 1.

RetrogradeA

i

Temperature

Such a diagram is produced for one particular mixture composition. Phase diagrams are deter-mined by use of equilibrium vapour ratios, K (or K values). The K value is defined as:

Yi

Ki = -

‘i

where

yi = mole fraction of component i in the vapour phasexi = mole fraction of component i in the liquid phase

The K value for any component is dependent on temperature and pressure, and also, at highpressures, on the fluid composition.

There are several terms used to define the location of various points on the phase envelope.

Cricondenbar

Maximum pressure at which liquid and vapour may co-exist in equilibrium (point N).


ISection 2. Physical Properties BP Multiphase Design Manual

Cricondentherm

Maximum temperature at which liquid and vapour may co-exist inequilibrium (point M).

Retrograde Region

The area inside the phase envelope where condensation of liquid occurs on lowering pressureor increasing temperature.

Quality Lines

Those lines showing constant gas volume weight or mol percentages. These lines intersect atthe critical point (C) and are essentially parallel to the bubble and dew point curves.

Line ABDE on the above diagram represents a retrograde condensation process occurring, forexample, in a gas-condensate transport line as pressure drops due to friction. Point A repre-sents the single phase fluid outside the phase envelope. As pressure is lowered, point B isreached where condensation begins. As pressure is lowered further, more liquid formsbecause of the change in the slope of the quality lines. The retrograde area is governed by theinflection points of such lines. As the process continues outside the retrograde area, less andless liquid forms until the dew point is reached (point E). Below E no liquid forms.

In a compositional model the predictions of gas and liquid physical properties are performedthrough use of an equation of state, EOS. Any equation correlating pressure (P), volume (V)and temperature (T) is known as an EOS. For an ideal gas the EOS is simply :

PV = n RT

where:

n = number of moles of gasR = Universal gas constant.

-To account for the non-ideality of most practical systems the above equation is modified toinclude various correlating constants. The Peng-Robinson (PR) EOS, for example, is given by:

RT amP =--

V - b V(V+b) + b(V-b)

where for any mixture :

b = i yi biyi = mole mole fraction of component ibi = empirical constant for component ia = & Cj yi yj (1 - BN,j) ai aiBN,, = empirically determined interaction parameter for the two components i and j.ai, a, = empirical constants for components i and j. These are a function of temperature.


BP Multiphase Design Manual Section 2. Physical Properties

A number of data generation methods are available in BP’s flowsheeting program, GENESIS.When running BP’s multiphase flow program MULTIFLO, with a compositional description ofthe fluids, the GENESIS module FLODAT is first run to produce the phase envelope andphysical property data. When running FLODAT the user is asked for two data generators.

1. The first is for K value determination, i.e. for generating vapour liquid equilibrium data.

2. The second is for the generation of thermal properties and densities.

In the absence of other recommendations, the most suitable data generators for typicalgas-condensate systems are:

1. [P-R (Peng-Robinson with interactive parameters) for K values

2. Z-J (Zudkvitch-Joffe) for thermal properties and densities.

issue 2 - December 1994 7

ISection 2. Physical PropeWes BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 2. Plysical Properties

2.3 Black Oil Models

The black oil approach to the prediction of phase behaviour ignores the fluid composition andsimply considers the mixture as consisting of a gas and liquid phase in which the gas may bedissolved in the liquid. The basic assumption of a black oil model is that increasing systempressure (and reducing temperature) cause more gas to dissolve in the liquid phases, and,conversely, decreasing system pressure (and increasing temperature) cause gas to evaporatefrom the liquid phase. It was noted in Section 2.2 that retrograde condensation involves theconversion of gas to liquid on reducing pressure. This is contrary to the fundamental assump-tion of the black oil model and so the black oil approach is only valid for systems operating atconditions far removed from the retrograde region.

Consider the diagrams below which illustrate a typical expansion process from a reservoir A toa separator B for:

ReservoirA

Figure 2. A gas reservoir

A Reservoir Critical

Temperature

Figure 3. A crude oil reservoir


Section 2. Physical Properties BP Multiphase Design Manual

For the liquid reservoir, lower diagram, the reservoir condition is well to the left of the criticalpoint and so the expansion process involves the continual evolution of gas, i.e. the operatingpoint moves steadily across the quality lines to a condition of ever decreasing liquid content.This type of process would be adequately represented by a black oil model.

For the gas reservoir, upper diagram, the reservoir condition lies to the right of the critical pointso that on expansion, (reducing pressure) the operating point moves across the quality lines toa condition of increasing liquid content, i.e. retrograde condensation. This process could notbe represented by a black oil model.

As a general guide a black oil model should be adequate for describing crude oil-gas systems,while a compositional model is necessary to describe wet-gas, gas-condensate and densephase systems.

The black oil model employs certain concepts and nomenclature which require definition.These are discussed briefly below:

2.3.1 Producing Gas Oil Ratio (GOR)

This is the quantity of gas evolved when reservoir fluids are flashed to stock tank conditions.The units are standard cubic feet of gas per stock tank barrel of oil (SCFLSTB) measured at14.7 psia and 60°F.

The GOR of a crude is obtained by experimental testing. However, the GOR will varydepending on how many flash stages are employed to get down to stock tank conditions. Thenormal convention is to calculate GOR from the sum of gas volumes evolved from a multistageflash procedure (normally this involves 2 or 3 flash stages). This more closely represents condi-tions in the field with the pressure and temperature conditions chosen for the first stage flashapproximating to conditions likely to be experienced in the first stage separator in the field.

2.3.2 Solution Gas Oil Ratio (Rs)

This is the quantity of gas dissolved in the oil at any temperature and pressure. It representsthe quantity of gas that would be evolved from the oil if its temperature and pressure werealtered to stock tank conditions, 14.7 psia and 60°F. Hence, by definition the Rs of stock tankoil is zero.

The Rs crude at its bubble point is equal to the producing GOR of the reservoir fluids.

The volume of free gas present at any pressure and temperature is the difference between theGOR and the Rs. The volume of free gas is corrected for pressure,temperature and compress-ibility to compute the actual in-situ volume of gas and hence superficial gas velocity. Rs can beevaluated from standard correlations such as Glaso or Standing. These correlations require asinput the oil and gas gravity and the pressure and temperature conditions.

IO Issue 2 - December 1994


Section 2. Plysical Properties

t

Producing gas oil ratiom-------w

Bubble Bubble --t Pressurepoint point

at 60 “F at T,

Producing gas oil ratio__-----

Free gas attemperature T,,

pressure P,

--t Pressure

Figure 4.

2.3.3 Volume Formation Factor (Bo)

The volume formation factor is the ratio of the volume occupied by oil at any pressure and tem-perature to the volume occupied at stock tank conditions. The units are pipeline barrels perstock tank barrel (pbbl/STB). The volume formation factor of stock tank oil is thus 1 .O. Throughuse of Bo the volume flowrate and density of the liquid phase can be calculated, (see Section1.1).

Standard correlations are available to compute Bo. These require as input the oil and gasdensity, the Rs of the liquid at the conditions of interest, and the pressure and temperature.

2.3.4 Live Oil Viscosity (Ov)

The viscosity of the oil in a 2-phase pipeline depends on the stock tank oil viscosity (dead oilviscosity), the solution gas oil ratio at the conditions of interest, and the pressure and tempera-ture. Correlations are available to compute the live oil viscosity.

Issue 2 - December 1994 I I


I 8, at bubble point

Pressure Bubble pointpressure

Figure 5.

The correlations available for Rs, Bo and Ov will yield approximate values only and wherelaboratory or field data is available, these should be used to adjust and tune these correlations.The way in which the correlations are tuned will depend on the quantity of field data available.

Details of the way in which field or laboratory data are used to tune the correlations is givenin Appendix 8C of the MULTILFO User Manual.

The minimum physical property information required to run a black oil model is:

(a) Stock tank oil gravity

There is often some confusion about the definition of gas gravity and hence uncertainty aboutthe value of this data item.

.m-

The majority of the correlations provided in MULTIFLO are based upon multistage 2-3 stages)separation, and the gas gravity used should always be the total gravity based upon theweighted average gravity from each stage, viz:

i = 1 ,n C(sgi*Gi)Total gas gravity =

i = 1 ,n CGi

where:

sgi = gravity of the ith separator stage off-gasGi = free gas GOR at the ith separator stagen = number of stages in the separator train with the final stage at stock tank conditions.



Section 2. Physical Properties

(c) Total producing GOR

This should be taken as the sum of the gas volumes evolved from each stage of a multistageflash.




LEFT BLANK

14 issue 2 - December 1994

HP ,Il~~ltiphast~ I>esipI .Ucillll~rl Section 3. Oil / Gas Pipeline Design

Section 3. Oil / Gas Pipeline Design

Design of Multiphase Crude Oil and Gas Pipeline Systems

3.1

3.2

Line Sizing and Pressure Drop

Temperature Drop and TemperatureRelated Problems

3.3

3.4

Flow Regimes

Slugging Flows

References

Appendices


Section 3. Oil / Gas Pipeline Design BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 3. Oil / Gas Pipeline Design

3 Design of Multiphase Crude Oil / Gas Pipeline Systems

The purpose of this section is to detail the various factors that have to be considered whendesigning multiphase crude oil/gas pipeline systems. Such pipelines may be onshore, such asat Wytch Farm and Prudhoe Bay, or offshore, such as the SE Forties inter-platform line, and theMagnus satellite well flowlines.

3.1 Line Sizing and Pressure Drop

The first stage of any design study is the specification of a line size to provide the requiredcapacity within the available pressure constraints. However, in addition to sizing a line for theavailable pressure drop, consideration must be given to ensuring that the flowing velocitiesprovide reasonable operating conditions.

A multiphase pipeline is generally used to transport total wellhead fluids (oil, gas and water)from a wellsite, or a minimum facilities platform (MFP), to a remote processing plant. The pres-sure at the processing plant is normally known, i.e. the separator pressure. There is generallysome information available on the pressure available at the wellhead, or WHFP, from a wellheadpressure study. Hence the pressure drop available for the pipeline is generally known. Notethat as the study progresses into detailed design, the wellhead pressure data may be refinedand may become available on a year-by-year basis. It may also be possible, and is desirable,to base pressure drop predictions on the bottom hole flowing pressure (BHFP) and the sepa-rator pressure as the fixed points in the system.

A line size has to be chosen which will yield a pressure drop equal to, or less than, that avail-able. The total pressure drop, APT, is given by:

AP, = AP,+ AP,,+ AP,,,

where:

AP, = pressure drop due to frictionAP,, = elevational pressure drop (hydrostatic head loss)AP, = pressure drop to acceleration of the fluids

AP, is normally insignificant and will not be discussed here. (The AP,, term should alwayscomputed in multiphase simulators such as MULTIFLO but only becomes significant for highvelocity flows approaching critical conditions).

3.1.1 Frictional Pressure Drop AP,

AP, is given by:

AP, = fo, LID p,, Vm*/2

where:

ftp = 2-phase friction factor


Section 3. Oil / Gas Pipehe Design BP Multiphase Design Manual

L = line segment length (m)D = line diameter (m)Pns = no-slip density (defined in Section 1 .l) (kg/m?

“m = mixture velocity (defined in Section 1.1) (m/s)

ftp is given by:

ftP = f”S mtp

where:

f”, = no slip friction factor (Darcy-Weisbach) based on homogeneous 2-phase physicalproperties and Reynolds number (see Section 1 .l).

mtP = a dimensionless 2-phase multiplier used to account for slip between the phases.

Most of the commonly used 2-phase pressure drop methods compute mtp as a function of thecalculated in-situ liauid hold-ub. H,, (for definition of H, see Section 1 .l). The Beggs and Brillmethod calculates mtp as follows:

mt, = es

where:

In (Y)s =

(-0.052 + 3.18 In (y) - 0.872 [In (y)12 + 0.0185 [In (y)]3

y = A,/ H,*

A, = no slip hold-up, (see Section 1 .l)

“,I + “sg

-

H, = liquid hold-up

The Beggs and Brill correlation for H, in a horizontal line gives:

a XLbH, = -

FrC

where:

Fr = Froude number, V,* / gD

a, b, and c are constants provided for each of the 3 flow regimes recognised by the Beggs &Brill method; segregated, intermittent, and distributed.



The Beggs and Brill method also has a term to correct hold-up for the effect of inclination:

where:

cp = is provided as a function of inclination (0) and flow regime.

Full details of the Beggs and Brill pressure drop and hold-up correlations, and other commonlyused correlations, are given in (1).

3.1.2 Elevational Pressure Drop (PeI)

Once the liquid hold-up, H,, has been evaluated the elevational pressure drop term is simplycalculated from :

*PC?, = g h b, H, + P,U -&II

where:

AP,, = elevational pressure drop (N/m*)g = acceleration due to gravity (9.81 m/s’)h = height increment (m)p, = in-situ liquid density (kg/m?p, = in-situ gas density (kg/m3)

3.1.3 Recommended Pressure Drop Methods

Multiphase design computer programs, such as MULTIFLO, generally provide a wide range ofthe most commonly used correlations for pressure loss and hold-up.

The Multiphase Flow group have tested the accuracy of these correlations using field datacollected from a number of sites, (2) (3) and (4).

The Beggs and Brill pressure drop method incorporating the Beggs and Brill hold-up correla-tions, was generally found to be the most reliable of the commonly used methods. It normallyover predicts pressure drop by between O-30%, and hence it provides a degree of conserva-tion.

However, for hilly terrain pipeline systems the situation is less clear. For the Wytch Farm in-fieldlines, the original Beggs and Brill method was found to under predict pressure loss. Studiesrevealed that this was due to an over-prediction of pressure recovery in the downhill sections,(3). A modified version of Beggs and Brill has been implemented in MULTIFLO which calculateselevational pressure drop in downward sloping sections assuming gas head recovery only (i.e. itneglects liquid head recovery). This modified version of Beggs and Brill gave very good agree-ment with data collected at Wytch Farm.

Data collected from the Cusiana Long Term Test (Lll), and the Kutubu hilly terrain pipelinesindicate that pressure recovery in downward sloping lines is a complex function of flow regime



and flowrate. A new computer model is currently being developed and tested to enable pres-sure loss to be more accurately evaluated for hilly terrain pipeline systems. The model tracksslug lengths through the pipeline and by taking account of slug decay in downhill sections it canprovide a more realistic estimate of the pressure recovery term. This development is discussedfurther in Appendix 3D.

Data collected on the Forties Echo - Forties Alpha (FE - FA) inter-platform multiphase linessuggested that the best correlations to use for flow in the risers are:

l Upflow Risers - Orkiszewski(generally over-predict pressure loss)

l Downflow Risers - homogeneous(tends to slightly under-predict pressure recovery)

In summary, the recommended pressure drop correlations for multiphase crude oil/gas linesare:

(a) Horizontal and near horizontal linesModified Beggs and Brill, gas head recovery only in downhill sections. For hilly terrainpipelines the most accurate determination of pressure loss will be obtained using the newBP model which is currently under development and test.

(b) Upflow Risers - Orkiszewski

(c) Downflow Risers - homogeneous

A recent study conducted by the National Multiphase Flow Database Project tested the accu-racy of the most commonly used pressure drop correlations against crude oil/gas field datasupplied by members, (5). It concluded that the Orkiszewski vertical flow correlation in combi-nation with the Beggs and Brill correlation for horizontal flow were the most consistently accu-rate pressure drop methods available. Hence apart from the hilly terrain pipeline considerations(and the National Database has little hilly pipeline data) the Database Project’s findings are inline with our own conclusions.

-

3.1.4 Flowing Velocities

Having selected a line size to meet throughput and pressure drop constraints, it is important tocheck whether acceptable flowing conditions exist within the pipeline.

The pressure gradient in a crude oil/gas multiphase line normally lies in the range 0.2-2.0psi/l 00 ft. However, a more useful means of assessing whether flowing conditions are accept-able is to check whether fluid velocities lie within certain limits.

At the.low end of the normal throughput range the actual (not superficial) liquid velocity, V,,should ideally be greater than ca. 3 ft/s. This will ensure that sand and water are continuouslytransported with the liquid phase and not allowed to drop out and accumulate at the bottom ofthe pipe. Details of the calculation procedure recommended to accurately assess the criticalsand transport velocity will be given in Section 12.

Note that the actual liquid velocity, V,, is given by:


BP Multiphase Design Manual Section 3. OU / Gas Pipeline Design

where:

V,, = superficial liquid velocityH, = liquid hold-up.

At the maximum throughput conditions the mixture velocity, V,,, should not exceed the erosionalvelocity limit, V,.

The current industry standard method of determining Ve is through use of the relationship giveninAPI 14E:

where:

V, = maximum acceptable mixture velocity to avoid excessive erOSiOn (fi/S)

Pm = no-slip mixture density (Ib/ft3)C = a constant, given in API 14E as 100 for carbon steel

This relationship is considered to provide an inadequate description of the erosion process.R&D is currently underway to provide a more satisfactory means of determining V,.

A review by the Materials and Inspection Engineering Group, ESS, advises that provided thatthe fluid does not have a high sand content, a value of C equal to 135 can be used for carbonsteel lines, (6).

Materials and Inspection Engineering Group, MIE, have recently issued a comprehensive guide-line on erosion in single and multiphase pipelines, (10).

In some systems flowing velocities need to be limited by a requirement to avoid removal ofcorrosion inhibitor. MIE recommend a maximum wall shear rate to avoid inhibitor stripping of100 N/m2, (11).

For duplex stainless steel lines a C factor of 236 is recommended. It is pointed out that thesemodified C factors were based on limited data and that critical areas of pipework should alwaysbe regularly monitored using NDT and corrosion probes as appropriate.



-


LEFT BLANK



Temperature Drop and Temperature Related Problems

3.2.1 General

In most multiphase simulators, such as MULTIFLO, heat transfer rates are calculated usinghomogeneous mixture properties. Hence no attempt is made to account for the position of theliquid and gas in the pipe. This approach has been found to be satisfactory for multiphasepipelines operating under normal transportation conditions. For example, good agreement wasfound between predicted and actual temperature drop for the FE-FA 6” and 12” flowlines.

In some circumstances the homogeneous heat transfer model may not be adequate. Hencewhen investigating heat flow from a pipe, and valves, to a multiphase fluid during a rupture,knowledge of the location of the liquid film may be of great importance. Under these circum-stances it would be necessary to use a sophisticated transient model.

3.2.2 High Temperature Related Problems

Transportation of high temperature wellhead fluids (in excess of ca. 90%) may give rise to prob-lems in meeting the maximum temperature specification of the pipe coating material.Approximate maximum temperatures for various pipe coatings are given below.

Coating Maximum Temperature (“C)

FBE (fusion bonded epoxy) 90

Coal Tar Enamel (coal tar wrap)

(i) on subsea lines(ii) on lines buried on land or at sea

7050

Neoprene ca. 100

The coating specialists in the Materials and Inspection Engineering Group should be consultedif there is any possibility of operating near the maximum allowable coating temperature. Thisgroup can also advise on alternative coatings for any particular application.

CO, corrosion of pipelines increases with temperature but reaches a maximum at about 60°C.In some cases it is necessary to cool the fluids prior to entry to the pipeline to reduce corrosionrates to a level consistent with the use of carbon steel pipework.

An alternative to providing cooling offshore is to use corrosion resistant pipe for the initialsections of the pipeline. At some distance downstream of the start of the line natural cooling inthe sea will reduce fluid temperatures to a level where carbon steel pipe can be used.

The Materials and Inspection Engineering Group, ESS, should be consulted for advice oncorrosion rates, corrosion inhibitor policy, and acceptable line inlet temperatures.


i


3.2.3 Low Temper&we Related Problems

The main areas of concern with regard to low temperature operation of multiphase lines are:

l Wax deposition

l Hydrate formation

l High viscosities, in some crudes, and under some circumstances, in oil/wateremulsions.

(a) Wax Deposition

In subsea and onshore multiphase pipelines, fluid temperatures may drop below the waxappearance point with the result that small particles of wax form. Some of these wax particlesare carried along in the liquid and some are deposited on the pipe walls. In fact because thewalls of the pipeline are generally colder than the bulk fluid, wax deposition onto the pipe wallmay occur while the bulk pipeline fluids temperature is still somewhat above the wax appear-ance point.

The pipe wall temperature is not printed in the MULTIFLO output, but can be simply calculatedin the following way.

For each pipe section MULTIFLO prints out the following heat transfer coefficients(Btu/hr “F ft2).

INTFLM - internal film coefficientEXTMED - external film coefficientINSLTN - heat transfer coefficient for the pipe coating or insulationPIPEWL - pipe wall heat transfer coefficientRADTN - contribution to heat transfer from radiationOVRALL - overall heat transfer coefficient

(All these heat transfer coefficients are based on the pipe internal diameter)

When radiation effects are negligible the overall heat transfer coefficient is given by:

1 1 1 1 1= + + +

OVRALL INTFLM EXTMED INSLTN PIPEWL

When radiation contributes to heat transfer, OVRALL is given by:

1OVRALL = + RADTN

(VINTFLM) + (1 /EXTMED) + (l/INSLTN) + (l/PIPEWL)

The heat flow through the pipe can be calculated from:

H = U A (Ti - TO)



where:

H = heat flow (Btu per hr per foot length of pipe)U = overall heat transfer coefficient, OVRALL (Btu/hr “F ft2, based on pipe id)A = surface area of pipe ID (ft2 per foot length of pipe)Ti = bulk temperature of fluid in pipe (“F)T, = bulk temperature of fluid external to pipe (“F)

Through any section of pipe:

H = hA(T,-T&

where:

h = heat transfer coefficient for layer or film (DCTMED, etc.)T, = temperature of one side of layerT, = temperature at other side of layer

If we draw a section of the pipe as:

Fluid Tiin pipe

Figure 1. Pipe cross-section

Then starting from outside of pipe, we have:

HT, = T, +

EXTMED A

\\\\\\\II

1

:

!

JJ/

Externalfluid To



where:

T, = temperature at outside surface of insulation

The temperature at inside surface of the pipe wall, T3, is given by:

HT, = Ti -

INTFLM A

It is important to be able to predict wax deposition rates as this may have a significant effect onpipeline operability. Wax deposition may increase to the point where it affects pressure drop. Ifwax is allowed to deposit in a line over a prolonged period intelligence pigging (for inspection)may become a risky operation because of the danger of blocking the line with a large plug ofwax. It is in fact possible to develop procedures to gradually remove the accumulated wax butthis can be a prolonged and costly exercise.

Wax appearance points are normally determined by the Production Operations Group (POB)Sunbury (7). POB can also determine wax deposition rates. The extent and nature of waxdeposits may dictate a need for regular pigging of a pipeline. It may also affect the choice ofhow production from a new field is linked into an existing system. For example, a requirementfor regular pigging may rule out the use of a subsea tee for a smaller line feeding an existinglarger one.

Listed below are wax appearance points for some typical stabilised crude& and one live crude.The wax appearance point is not generally available for live crudes (although it can be obtainedby POB) but it is always lower than that of a stabilised crude.

CrudeWax Appearance Point

Stabilised Crude (“C) LIVE Crude (“C)

Forties 32 24Andrew 28Clair (low wax sample) 11Clair (high wax sample) 17Ivanhoe 36

It should be noted that although some systems operate below the wax appearance point, theirwax deposition rate may be so low that there is no requirement for regular pigging.

Further R&D effort is required to understand many aspects of wax deposition in multiphasesystems. A major JIP is currently being set-up to build and collect data from a multiphase rigoperating under conditions where wax deposition occurs. The ultimate aim of the JIP will be thedevelopment of models to predict wax formation and deposition characteristics in multiphasepipelines.

Wax crystal modifier chemicals are used in some systems to reduce was deposition and henceavoid pipeline operating problems. These chemicals do not prevent waxes forming, but alterthe crystal structure to make it more difficult for the wax to adhere to the pipe wall. The waxparticles are then more likely to be carried along with the liquid.



Problems relating to wax deposition should be taken up with PO6 Sunbury at an early stage ofany project so that a testing programme can be planned and carried out in time to answer keyquestions relating to design and operability.

(b) Hydrate Formation

Hydrates are ice like materials which can form under certain conditions of temperature andpressure when water is present. Hydrates formation is promoted by high pressure and lowtemperature. The formation of hydrates is undesirable because they can obstruct the flow,causing an excessive pressure drop, and, if dislodged, can cause impact damage at bends orin vessels.

Hydrate formation conditions can be determined by the GENESIS program, or looked up in theGPSA charts.

The prevention of hydrate formation in multiphase systems can be accomplished by one of thefollowing means :

l Maintain system pressure below the hydrate point.

l Maintain system temperature above the hydrate formation point.

l Remove all water from the hydrocarbons(this would normally require the gaseous phase to be dehydrated).

l Injection of conventional hydrate inhibitors.

l Injection of threshold hydrate inhibitor.

A brief summary of the advantages and disadvantages of the various hydrate inhibitors is givenbelow.

(bl) Methanol

Methanol has the advantage over the glycols (MEG, DEG and TEG) of being able to dissolveformed hydrates. It also has a lower viscosity, and hence better dispersion characteristics thanthe glycols. Methanol is, however, more hazardous because of its higher vapour pressure andflash point.

If the gas stream from a methanol dosed multiphase stream is treated in a glycol contactordehydration plant (normally using TEG) some of the methanol vapour may be absorbed in thecontactor increasing &oiler duty and loading in the still column. The absorbed methanolvapour will then pass with the water vapour from the column and be vented to atmosphere.

($2) Monoethylene Glycol (MEG)

MEG will not dissolve formed hydrates. If gas from an MEG dosed stream passes to a glycolcontactor at the reception plant, there is a potential problem of MEG degradation. The degra-dation and boiling point temperatures of MEG are 165°C and 197°C respectively, compared toa typical TEG contactor reboiler operating temperature of ca. 204°C.



(b3) Diethylene Glycol (DEG)

As for MEG but vapour pressure is significantly lower (hence less passes to contactor). Thedegradation and boiling point temperatures of DEG are 164oC and 245oC respectively. Themain disadvantages of DEG is its availability and cost. Also, DEG has a higher viscosity andinferior dispersion characteristics in comparison to MEG.

(b4)Triethylene Glycol (TEG)

There are no compatibility problems with the glycol contactor unit. The main problem with TEGis its high viscosity and hence poor dispersion characteristics.

(b5) Threshold Hydrate Inhibitor

The conventional alcohol and glycol inhibitors discussed above work by altering the thermody-namic phase equilibrium so that hydrate formation occurs at a reduced temperature. Thesechemicals are typically injected at rates of 1 O-40% by volume on the water phase. Recently anew type of inhibitor has been developed which is effective at very much lower dosage rates.These threshold hydrate inhibitors (THI) do not affect the phase equilibria but alter the kinetics ofthe hydrate formation process so that the possibility of hydrate formation occurring within thepipeline is reduced to a very low level.

Field and laboratory tests indicate that THI will be effective at concentrations of between 1000and 5000 ppm. A THI field trial was conducted on the Ravenspurn to Cleeton gas/condensateline in May 1994. The trial showed conclusively that THI prevented hydrate formation for 6 dayswhen it was used in place of methanol. Prior to injection of the THI a blank test was carried outduring which neither THI nor methanol was injected. There was clear evidence that hydratesformed and partially blocked the line during this period.

(c) High Viscosity Fluids

As the temperature in a multiphase pipeline reduces so the viscosity of the oil phase increasesand this leads to an increased frictional pressure gradient. The viscosity of stabilised and live oilcan be determined by POB, Sunbury, as a function of temperature, and these values can beused to tune the viscosity correlations in MULTIFLO.

When oil/water emulsions are present determination of the liquid viscosity is far more compli-cated. The viscosity of an emulsion depends on the particle size distribution, which is largelydetermined by the flow history of the fluid and the shear imparted by chokes, pumps, etc.Dispersions that have undergone high shear (e.g. by flowing through a high pressure dropchoke) will contain water particles of a relatively small size and this so called “tight” emulsion willhave a relatively large viscosity.

Once formed the emulsion will generally exhibit non- Newtonian shear thinning flow behaviourso that its viscosity is dependent on the shear rate in the pipeline, i.e. the prevailing flowrate.

Emulsion viscosities are thus dependent on:

l Temperature

l Water cut


BP Multiphase Design Manual Section 3. Oil/ Gas Pipeline Design

l Flow history, including shear rate experienced when formed

l Flowing shear rate, i.e. flowrate in the pipeline.

In MULTIFLO emulsion viscosities can be estimated using a weighted average of the oil andwater. Alternatively the Richardson emulsion viscosity correlation can be used. This gives theuser the ability to assess the sensitivity of the pressure drop predictions to various possibleemulsion viscosities by altering the constant in the correlation.

When experimental data on emulsion viscosities become available, the constants in theRichardson correlation can be varied so that predicted emulsion viscosities approximatelymatch the actual data for the range of flowrates (shear rates) that will be experienced in thepipeline.

The effect of viscosity on flow regimes and slug characteristics is discussed in Section 3.3.2below.

3.2.4 Thermal Insulation of Pipelines and Risers

In some circumstances it is practical and cost effective to thermally insulate a pipeline system.Insulation may be provided for any one, or a combination of, the following reasons:

l Avoid or limit wax deposition.

l Avoid or limit hydrate formation.

l Minimise oil emulsion viscosity to avoid excessive pressure drop.

l Maintain fluids at a sufficiently high temperature to avoid, or limit requirement for, feedheating at the separation plant. A certain minimum separator temperature may be necessaryto ensure satisfactory oil/water separation. A minimum temperature may also be required tomeet the oil vapour pressure specification.

Pipelines and risers are always provided with some form of coating for corrosion protection andthese offer varying degrees of thermal insulation. Risers on offshore platforms are normallycoated with coal tar enamel or a very thin layer of fusion bonded epoxy (FBE). These offer littlethermal insulation. Often risers are coated with ca. 12 mm of Neoprene to provide corrosionprotection in the splash zone. Where thermal insulation is required this coating may be usedover the full length of the riser.

Pipelines are normally protected against corrosion by FBE or coal tar enamel. For an offshorepipeline requiring concrete weight coating (to provide negative buoyancy) coal tar enamel isnormally applied to the pipe to provide corrosion protection.

Offshore pipelines smaller than 16 ins. in diameter are generally required to be trenched toavoid being damaged by fishing trawl boards. Even if a pipeline trench is not mechanicallybackfilled (i.e. not buried) some natural in-fill of the trench normally occurs so that the pipebecomes partially buried. A survey of the Forties 32” MOL showed that most of the pipe hadbecome totally covered in soil. In MULTIFLO the user can model partial burial by specifying thefraction of pipe surface exposed to the sea. The default value is 1 .O, i.e. fully exposed pipe.



Under some conditions a trenched pipeline has to be covered in rocks to provide stability or toprotect it from upheaval buckling. This so called rock dumping buries the line providingenhanced thermal insulation. On some occasions pipelines are buried solely to provideenhanced thermal insulation. The FE-FA pipelines were buried for this reason.

-

In addition to the coatings referred to above, low thermal conductivity insulation materials canbe applied to minimise heat loss from a pipeline. Insulation systems can be divided into 2 maincategories:

(a) Low strength polymeric based materials requiring protection from theenvironment by encasing in an outer sleeve.

Examples of such system are:

l Central Cormorant multiphase production lines, which are insulated with polyurethanefoam (PUF) encased in an outer steel sleeve. (Note that pipeline coaters are now offering asystem for subsea use consisting of PVC foam sandwiched between an ethylene propylenediene monomer (EPDM) corrosion coating and an EPDM outer layer providing protectionfrom the environment).

l Miller Landline. This is insulated with PUF encased in an outer layer of polyurethane.

(b) Insulation capable of withstanding the environment.

Examples are:

l Neoprene, as used on the FE-FA flowlines.

l EPDM, as used on the Don-Thistle flowline.

l Syntactic materials, such as syntactic PUF as used on Central Brae.

For further details on pipeline insulation materials available, costs and installation, pleasecontact the Pipeline Group of XFE, BPX (Ian Parker).

Listed below are the thermal conductivity of the coatings and insulation materials previouslyreferred to:

Material Thermal ConductivityBtu (hr “F ft) W b-W

Concrete (saturated low density)Concrete (saturated high density)FBECoal Tar EnamelNeoprene/EPDMPolyurethane Foam (PUF)PVC FoamSyntactic PU

1.66 2.871.18(l) 2.040.114 0.20.114 0.20.156(2) 0270.017(3) 0.030.023 - 0.04(3) 0.04 - 0.70.07(3) 0.12

I6 Issue 2 - December 1994

rBP Multiphase Design Manual Section 3. Oil / Gas Pipeline Design

Notes

(1) Most pipelines now use high density concrete.(2) Ageing tests on Neoprene carried out in 1985 showed that the thermal conductivity rises to

ca. 0.38 W/mK as water absorption occurs.(3) Values depend on water depth, temperature and age.




LEFT BLANK

-.



3.3 Flow Regimes

3.3.1 Discussion

Detailed descriptions of the various flow regimes occurring in horizontal and vertical flow weregiven in Section 1.2 of this manual.

50Dispersed

10 -

$- l-52

50.1 -

0.01 -

z 1

50.1

0.01

Horizontal flow

0.1 1 10 100 500 U% (m/s)

Dispersedb u b b l e ( D B ) -

Intermittent (I)

Annular

Stratifiedwavy (SW

Downflow

0.1 1 10 100 500 USC (m/s)

Upflow

0.1 1 10 100 500 usG (m/s)

Figure 2. Taitel-Dukler Flowmaps



The transition between such flow regimes is usually presented in the form of a two dimensionalflowmap. The usual co-ordinates of such a map are superficial liquid velocity (effectively liquidvolume flowrate) and superficial gas velocity. The early flowmaps were derived from flow obser-vations. These empirically based flowmaps are not particularly accurate for systems where thefluid properties, pipe size and inclination are different from those for which the flowmap wasoriginally produced.

More recently flowmaps have been developed which incorporate physical models for each tran-sition process. Models of this type, such as that of Taitel-Dukler (T-D), attempt to account forthe effects of pipe diameter, pipe inclination and gas and liquid density and viscosity. In Figure2 flowmaps are presented for a horizontal pipe, one with a small downhill inclination, and onewith a small upward inclination. These flowmaps, which were produced using the T-D proce-dure, show how small inclinations significantly affect the position of the flow regime transitions.For a downward sloping line the area of the stratified flow regime increases. For an upwardlyinclined line the stratified flow region shrinks and as the slope is further increased, disappearsentirely.

One of the main reasons for producing a flowmap is to determine the conditions under whichslug flow occurs. It can be seen in Figure 2 that the slug flow region occupies a fairly centralposition on the map, i.e. slug flow occurs at moderate gas and liquid flowrates. This type ofslug flow is commonly encountered in pipelines carrying live crude oil such as the Magnus andBuchan satellite well flowlines, the Forties Echo to Alpha in-field flowlines, and the Wytch Farmin-field lines.

3.3.2 Effect of Viscosity on Flow Regimes

Experimental work has been performed at Sunbury to investigate the effect of viscosity on flowregimes and slug characteristics. The studies, performed on a 2” air - water/glycerol rig,showed that the slug flow region of the flowmap expands with increasing viscosity. Thisincrease in the area of the slug flow region is in close agreement with the predictions of theTaitel-Dukler flowmap.

No data has been collected on the effect of viscosity on slug characteristics. However, the BPmodel correctly predicts a higher equilibrium liquid hold-up with increasing viscosity. As a higherhold-up in practice results in more frequent slugging, we can confidently predict that increasingliquid viscosity will give rise to smaller, more frequent slugs.

3.3.3 Recommended Flowmap

In the BPE Report “Summary of Field Data on Slugging Multiphase Systems” (8), it wascommented that flow regime data collected by BP had shown that the original (T-D) flowmapwas “unsafe” in that in many instances it predicted the flow to be annular when in fact slug flowwas observed. The Barnea modified T-D flowmap predicts that the annular to slug transitionoccurs at higher gas velocities than the original T-D and so has a larger slug flow region. All thetests where slug flow was observed at Prudhoe Bay were predicted to be in slug flow by theBarnea modified T-D flowmap. On some occasions, however, annular flow was experiencedwhen slug flow was predicted. The Barnea modified T-D flowmap can thus be thought of asproviding a safe, or conservative, prediction of flow regime.

The Barnea modified T-D flowmap is now available in MULTIFLO and is the recommendedmethod of evaluating flow regime in crude oil/gas multiphase pipelines.

.-

2 0 Issue 2 - December 1994


3.3.4 Effect of Pipeline Geometry and Risers on Flow Regime

The type of slug flow indicated by flow regime maps arises as a result of fluid dynamic insta-bility. It is frequently described as normal slug flow since it can occur in perfectly horizontallines and does not depend for its generation on line topography or throughput changes.

Slugs may, however, be generated by other mechanisms. In particular when stratified flowoccurs in a pipeline which undergoes changes in inclination, liquid may accumulate at lowpoints to form a temporary blockage. Eventually the pressure builds up behind the blockagecausing the liquid to be expelled and transported through the line as a slug. This type of flow istermed terrain, or geometry dependent slug flow.

when a flowline terminates in a riser a particular type of terrain slugging may occur wheresome, or all, of the riser fills with liquid. This phenomenon, known as severe slugging, impartsgreater perturbations on the process plant, and larger forces on pipe bends, than does normalslug flow. Severe slugging occurs when stratified flow occurs in the flowline, and the ratio ofgas to liquid flowrate is below some critical value such that the rate of pressure build-upupstream of the blockage at the riser base is insufficient to overcome the increase in hydro-static head as the slug grows in the riser. (Severe slugging is discussed in detail in Section3.4.2 and in Appendix 3C).

The conditions leading to the occurrence of severe slugging have been determined from experi-mental tests at Sunbury.

The Multiphase Flow Group at Sunbury have developed a mechanistic model which describesthe conditions leading to the occurrence of severe slugging. This model has been validatedwith data obtained from the Sunbury experimental facility. The mathematical expression devel-oped defines a critical liquid velocity above which severe slugging will occur:

[L (P, - P,) g (1 - HLl sin 4

where:

P,, = separation pressure (N/m?L = line length upstream of the riser(m)p, = liquid density (kg/m3)p, = gas density (kg/m3)H, = average liquid hold-up in lineg = acceleration due to gravity (9.81 m@)0 = inclination of riser to horizontal (i.e. sin f3 is usually 1)

This expression, in a slightly different form, was later published by Shell following studies carriedout on an experimental facility at the KSLA Laboratories. Shell defined a dimensionless numbernTT,, as:

“s, p

37 =ss

Ps, 9 L PL (1 - WI


iSection 3. Oil/ Gas Pipeline Design BP Multiphase Design Manual

Hence when ~~ > 1, severe slugging is not experienced.

The expression developed by BP for severe slugging can be drawn on the T-D flowmap inMULTIFLO, and this then provides a flow regime map for the complete flowline-riser system.When the flow regime is predicted to be stratified-wavy, but lies to the right of the severe slug-ging line (i.e. at gas velocities in excess of that required for severe slugging), then wavy flow willoccur in the flowline, and a churn or annular type flow will be experienced in the riser.

When the flow regime is anything other than stratified or stratified-wavy flow in the flowline, thenthat type of flow will persist through the riser. In particular if normal slug flow occurs in the flow-line, then the slugs will continue on through the riser. Field data gathered from the FortiesEcho-Alpha inter-field pipelines have shown that at moderate flowrates, slugs, once formed inthe flowline, do persist essentially unaltered through the riser. However, at lower flowrates itwas observed that although slug flow was maintained through the riser, the slugs underwentsignificant deceleration in the riser. This type of flow, termed slug deceleration, is discussedfurther in Section 3.4.1 below.

-



3.4 Slugging Flows

3.4.1 Normal or Hydrodynamic Slugging

It was discussed in section 3.3 that normal slug flow occurs at moderate gas and liquidflowrates and hence is commonly encountered in multiphase pipelines. The intermittent natureof this type of flow means that the separation plant at the downstream end of a slugging multi-phase line will experience variations in liquid and gas flow.

In order to be able to design reception facilities to accommodate slug flow, and to design pipesupports to handle the forces associated with slug flow, it is necessary to predict both of thefollowing:

l Slug volume (slug length and liquid content)

l Slug velocity

(a) Slug Volume (Slug Length and Hold-up)

A considerable amount of R&D work has been carried out world-wide to improve the under-standing of slug flow. It was apparent from early experimental work that slug lengths observedin small scale laboratory rigs could not be scaled up to field conditions.

In the late 1970s the operators of the Prudhoe Bay Field, Alaska, (PBU) were faced with theproblem of designing numerous large diameter multiphase flowlines (24” in diameter) and theseparators into which the fluids pass. They commissioned fluid flow specialists at TulsaUniversity (Dr. J.P. Brill) to carry out a large programme of work which included the collection ofslug flow data from existing lines at PBU. The largest of these was a 16” 3-mile pipeline. Fromthe data collected at PBU, together with some information on smaller flowlines, and laboratoryscale multiphase systems, a correlation was developed to predict mean and maximum sluglength. This correlation was published by Brill et al in 1979 :

In(L,) = -2.663 + 5.441 [ln(d)P5 + 0.059 [ln(V,)]

where:

L, = mean slug length (ft)d = pipe diameter (in)V, = mixture velocity (ft/sec)

The PBU data suggested that slug lengths follow a log normal distribution, i.e. slugs distributedabout the mean length through the following relationship:

In(L,) = G* zp + In(L,)

where:

L, = slug length observation



(T = standard deviation (approximately 0.5)zp = standard normal distribution function (3.08 for 0.1% probability)

Based on the 16” PBU data, Brill et al claimed that the longest slug likely to be observedequates approximately to the 0.1% probability slug. The calculated 0.1% probability slug isapproximately 4.7 times longer than the calculated mean slug.

The above slug length correlation has become the industry standard design method and hasbeen used extensively by oil companies and contractors since 1979.

In 1980/81 the Tulsa University workers returned to PBU to gather data on a new 24” line thatran parallel to the 16” line that had been monitored previously. They found that in general slugswere not as large as had been predicted using their original correlation. BP revised the originalslug length correlation to take account of the new 24” data. It is this modified correlation whichis available in MULTIFLO. It is referred to as the Brill method as the correlation was kept in thesame form as Brill’s original equation:

In(L,) = -3.579 + 7.075 [ln(d)]0.5 + 0.059 [In (V,,,)] - 0.7712 [In(d)]

The above correlation is heavily biased by the PBU data. Most of this data was collected on a16” and 24” line which run parallel to one another and hence have the same length and geom-etry. The correlation thus shows no dependence on line geometry or fluid type, and only a veryweak dependence on flowrate. It was widely felt that the correlation could prove unreliablewhen applied to other types of systems. Consequently, under BPX, XTC, sponsorship theMultiphase Flow Group commenced a programme of work aimed at gathering slug flow dataon a wide range of multiphase systems.

Since 1985 the Multiphase Flow Group have collected data from the following sites.

Wytch FarmWytch FarmMagnus Satellite Well FlowlinesForties Echo-Alpha in-field linesPrudhoe Bay WOAPrudhoe Bay WOA and EOAKuparukDon-Thistle

(Sherwood) 6 and 10”(Bridport) 4”

6”6” and 12”

(1987) 24”(1989) 24”(1990) 1 O-24”

8”

A summary of the data collected on all these tests, apart from the last 2, is available in (8).

The data collected by BP covers a wide range of pipe sizes, lengths, geometries, fluid proper-ties, and water cuts. The data has been compiled into a database vastly greater than anythingpreviously available. Using this database, the Multiphase Flow Group have developed a semi-mechanistic correlation for slug frequency. By applying one of the available slug flow models,knowledge of slug frequency yields the mean slug length. This is discussed in Appendix 3A.

No data has yet been collected from slugging hilly terrain pipelines. A new slug model hasrecently been developed by the Multiphase Flow Group to track slug sizes throughout a hillyterrain pipeline system. This model was referred to in Section 3.1.3, and is discussed further inAppendix 3D. It is intended to gather data from the Cusiana in-field flowline system. The datacollected from this hilly terrain system will be used to validate and develop the new model.


iBP Multiphase Design Manual Section 3. Oil / Gas Pipeline Design

In parallel with this data gathering exercise, the Multiphase Flow Group have carried out anextensive programme of work to determine how slug lengths are distributed about the mean.This work has involved analysis of data collected on experimental rigs at Sunbury as well asdata collected from other experimental facilities. Using the limited PBU data Brill et al hadconcluded that slug lengths follow a log-normal distribution. Hence the Brill correlation alwayscalculates the maximum slug length to be 4.7 times the mean. The Multiphase Group havefound that the distribution of slug lengths actually varies with the location of the pipeline oper-ating point on the flowmap. As a consequence the ratio of maximum to mean slug lengthvaries with flowing conditions. Near the stratified wavy boundary the maximum slug length maybe ca. 4-5 times the mean. However, near the elongated bubble transition maximum sluglengths are only ca. 2 times the mean. Work on quantifying slug length distribution is contin-uing at Sunbury.

The combination of a method for determining slug frequency, a slug flow model, and a sluglength distribution model provides a means of predicting mean and maximum slug lengths.This new approach is available in MULTIFLO as the “RCS Mechanistic Model”.

A more detailed explanation of the slug model is given in Appendix 3A.

Prediction of the liquid volume associated with slug flow requires knowledge of the liquidcontent of the slugs (slug hold-up), as well as the slug length. The slug hold-up methodcurrently used within the RCS slugging model is that due to Gregory. The Gregory model isbased on laboratory scale experiments and predicts slug hold-up as a function of mixturevelocity only. BP has found that in practice slug hold-up is strongly dependent on water cut aswell as mixture velocity. Hence at Prudhoe Bay slug hold-ups have increased from ca. 0.3 to0.9 as water cuts have risen from 0 to 50%. Similar significant increases in slug hold-up havebeen observed on the FE-FA 6” test line when running wells of different water cuts. Furtherwork is being conducted in this area in order to develop a more reliable design method.

(b) Slug Velocity and Forces due to Slugging

(bl) Horizontal Ihes

When slug flow occurs in an essentially horizontal line the mean velocity of the liquid in thebody of the slug is equal to the mixture velocity, V,, hence :

vs = mean slug velocity

It is this velocity, V,, which should be used when evaluating the forces imposed by a slug as ittravels through a bend.

A detailed discussion of how to calculate loads associated with slug flow is given in Section 11of this manual.

($2) Pipeline Risers

It was mentioned in Section 3.3 that although a slug will progress through a riser at the down-stream end of a pipeline, it will tend to decelerate as the line upstream of the slug packs up toprovide the pressure to overcome the increasing hydrostatic head in the riser.


ISection 3. Oil / Gas Pipeline Design BP Multiphase Design Manual

As the slug leaves the riser the hydrostatic head loss in the riser reduces so that the upstreamgas bubble expands and accelerates the slug into the process plant. This phenomenon isshown schematically in Figure 3.

Velocityof slug

Time

Figure 3. Normal slugging in flowline riser

Press

In order to model the effect of slug flow in a pipeline- riser, BPX has developed a dynamicprogram known as NORMSL. This can be used to determine the velocity of the slugs as theypass through the riser and into the topsides pipework. In order to assess the effect of sluggingon the topsides process plant the output file from NORMSL can be used as the input to adynamic simulation model of the topsides plant.

Process dynamic simulation work in BPX is often performed using the SPEED-UP program.(This type of work is performed by the Process Simulation Group of ESS, Sunbury). NORMSLand SPEED-UP have now been directly linked so that pressure changes in the process plantarising from the production of slugs or gas bubbles are fed back directly into the pipeline-risermodel. This feedback effect becomes more significant as the height of the riser increases, i.e.for deep water developments.



3.4.2 TerraWGeometry Dependent Slugging

(a) Hilly Terrain Pipelines

It was explained in Section 3.3 that when stratified flow occurs in a pipeline liquid may accumu-late at low points to form a temporary blockage. Gas pressure builds up behind the blockagecausing the liquid to be expelled as a slug. Such terrain.induced slugging is clearly dependenton the geometry of the pipeline as well as the flowing conditions.

As the available design methods for this phenomenon were inadequate and known to signifi-cantly overpredict slug size, the Multiphase Flow Group carried out physical and theoreticalmodelling studies to gain a better understanding of the terrain slugging process. These studieshave led to the development of a mechanistic model for slug production from dips. The modelhas been validated with data obtained from the experimental facilities. The new model isdiscussed in detail in Appendix 3B.

The two most important aspects of the model with regard to pipeline design are:

(al) Critical Gas Velocity

The critical gas velocity above which no liquid accumulates in a dip is evaluated by consideringthe gas velocity for total liquid removal by the co-current liquid film process.

(a2) Slug Size and Frequency

At gas velocities below the critical value liquid can accumulate in a dip and eventually this leadsto the production of a slug. As a slug moves through the uphill section, liquid is shed from itsrear and runs back down the slope. If insufficient liquid is available in the preceding film for theslug to scoop up and replace that lost by shedding, then the slug will collapse before reachingthe brow of the hill. For a system with a steady liquid inflow, liquid will build up in the dip so thateventually slugs will emerge to pass into the downstream pipework. The model evaluates thefrequency of slug production for a particular geometry. With knowledge of the slug frequency,the slug size is determined by calculating the volume of liquid entering the system during theinter-slugging period.

(b) Pipeline-Riser Systems

When a pipeline terminates in a riser a particular type of terrain slugging may occur which hasbeen variously termed riser and severe slugging. The conditions giving rise to the occurrence ofsevere slugging have been outlined in Section 3.3.

A diagrammatic representation of severe slugging is given in Figure 4. A detailed description ofthe severe slugging phenomenon plus details of the BP test rig studies is given in Appendix 3C.A brief description of severe slugging is given below:

When a stratified flow occurs in a pipeline and the ratio of gas to liquid flowrate is below somecritical value, a liquid blockage will form at the base of the riser. When such a blockage occursliquid accumulates at the base of the riser while gas is trapped within the flowline. The liquidslug now formed at the junction between the flowline and the riser will continue to grow if the


r-Section 3. Oil / Gas Pipeline Design BP Multiphase Design Manual

rate of hydrostatic head increase in the riser, corresponding to the rate at which liquid arrivesfrom the flowline, is greater than the rate of gas pressure increase in the flowline. Liquid accu-mulation continues until the riser is full of liquid (slug generation).

Stage 1Slug generation Start of Stage 1

Blockage occursat base of riser

Stage 2Slug production

Gas bubblereaches baseof riser

Stage 3Bubble penetration

Stage 4Gas blowdown

End of Stage 2Bubble starts topenetrate riser

- Slug productionto separators

End of Stage 3Gas reachestop of riser

--) Gas productionto separators

Figure 4. Severe slugging in flowline riser



At this point the hydrostatic head loss over the riser reaches a maximum value and the slugbegins to be pushed slowly from the flowline (slug production).

Once the gas slug interface enters the riser the hydrostatic head decreases rapidly while theexpanding gas bubble accelerates the bulk of the liquid from the system (bubble penetration).

This stage continues until the gas bubble enters the separator. Gas production rapidly rises toa maximum value and then declines steadily as the line depressures (gas blowdown).

As the gas flowrate reduces, any liquid held up in the riser falls back and accumulates togetherwith liquid arriving from the flowline, to form a blockage at the base of the riser. The cycle isthen repeated.

The occurrence of severe slugging is generally associated with flowlines which slope down-wards to the base of the riser. However, a similar surging cycle can be produced in horizontaland slightly upwardly inclined pipelines.

($2) Modelling

Extensive physical modelling has been carried out by the Multiphase Flow Group to investigateand confirm previous descriptions of severe slugging and to investigate means of eliminatingthe phenomenon, see Appendix 3C. Much of this work was sponsored by the SE Fortiesproject.

Of particular interest to the SE Forties Project was the use of riser gas injection to eliminatesevere slugging both at low throughputs and at start-up. Tests showed that gas injectionwould reduce the severity of the severe slugging cycle and that in sufficient quantity it wouldcompletely eliminate the phenomenon.

A computer model of severe slugging was purchased from the Tulsa University Fluid flowProjects (TUFFP). This program evaluates liquid and gas production rates throughout thesevere slugging cycle and so by feeding the results directly into a dynamic model of thetopsides plant, the effects of slugging on the process facilities can be evaluated. The TUFFPmodel was modified at Sunbury to include topsides pipework. The program could then beused to evaluate the forces exerted on the riser and topsides pipework. The program wasfurther modified to include the effect of riser gas injection. The predicted quantities of gasrequired to eliminate severe slugging were found to be in good agreement with the datacollected from the BPX Sunbury experimental facility.

It was found that the Taitel-Dukler-Barnea criterion for the onset of annular flow provided agood estimate of the quantity of gas required to eliminate severe slugging. This methodpredicts that the transition to annular flow occurs at superficial gas velocities in excess of a crit-ical value given by:

3.1 1s 9 0, - PJV4v 2w 44 0.5

where:

s = surface tension (N/m)g = 9.81 (m/sec2)



p, = liquid density (kg/m3)pg = gas density (kg/m?

The modelling work showed that partial closure of a choke positioned near the top of the riserwould also eliminate severe slugging.

The choke adds sufficient frictional pressure drop so that the system pressure loss becomesdominated by friction, rather than by hydrostatic head loss as is the case for the unchokedpipeline-riser. In order to establish a stable, friction dominated system, the choke needs toprovide a pressure drop comparable to the hydrostatic head loss over the riser when full ofliquid.

(b3) Operational Experience

(b3.1) General

The SE Forties field is the principal BP-operated site where severe slugging was considered inthe design stage. A number of the Magnus satellite wells may also have exhibited severe slug-ging had their maximum production rates been reduced to low values (their arrangement isdifferent to SE Forties in that the chokes are at the top of the riser - severe slugging would notoccur unless the riser- top chokes were almost fully open).

There is one 12” line and one 6” line between Forties Echo minimum facilities platform andForties Alpha. There are eleven wells on Forties Echo, now mainly assisted by electricsubmersible pump. They have chokes on Forties Echo. Therefore once the flows from thewells are co-mingled into the 12” line (or flowing individually or co-mingled through the 6” line),there is the possibility of severe slugging at low flowrate.

Low flowrate considerations obviously may also include start-up, when one or more severeslugging cycles may be experienced before a higher steady-state flowrate is reached.

In April 1988 a major field test was undertaken on the Forties Alpha platform mainly to investi-gate the behaviour of the 6” line over a range of flowrates, and from different wells that had arange of water cut. The aim was to keep reducing the flowrate from one well at a time throughthe 6” line until the onset of severe slugging. However, due to a fear that choking back thewells too far might kill them, it was not possible to reduce the production rates sufficiently toreach genuine severe slugging during the tests. Despite this limitation, there was a markedchange in the character of the flow pattern, from steady low liquid hold-up slugging to a regularsurging, as the flowrates were reduced.

Experience with the 12” line is limited to start-ups. Operators present at first oil through the linereported heavy surging. During the 1988 tests three start-ups were monitored, with severallarge surges during each. However, because of the steadily increasing input flowrates it wasnot possible to assess the accuracy of the predictive tools, either for the occurrence of severeslugging, or for the cycle time.

(b3.2) Elimination of Severe Slugging

The Forties Echo-Forties Alpha line has two methods for ameliorating the effects of severe slug-ging - riser-base gas injection, and riser-top choking. The theory behind these 2 methods isdiscussed in Appendix 3C.



The design intention for SE Forties was that riser gas injection would be the principal means ofeliminating or ameliorating severe slugging at low production rates and at start-up.

Gas is re-cycled after NGL extraction through a 3” line down to the base of the 12” riser, and a2” line for the 6” riser. The gas injection system capacity is sufficient to put the risers intoannular-mist flow at the highest gas rates.

As mentioned above it was not possible to undertake any ‘steady-state’ severe slugging testsduring the 1988 field work. However, the use of gas injection to ameliorate the low flowrateperiod of a line start-up was demonstrated on both diameter lines.

The full gas injection procedure was used during the third of the 1988 test sequence of threestart-ups of the 12” line, with the maximum recommended gas injection rate (180,000Sm3/day) into the 12” line flowing before the first production fluids were introduced at the Echoplatform. Whilst there was still an initial surging the gas injection did serve to reduce the dura-tion of the main surge, and also the number of cycles.

This benefit of gas injection during start-up was also demonstrated whilst trying to bring on aweak well through the 6” test line. The well began to flow, but as liquid reached the base of theriser at Forties Alpha the additional back pressure due to the increasing liquid head was suffi-cient to cause the well to stop flowing. To overcome this the gas injection system wasswitched on in the 6” line (up to 60,000 Sm3/day). The liquid head build-up was greatlyreduced, thereby minimising any back pressure and allowing the well to build up to itsmaximum production rate.

The Forties Echo risers at FA were also fitted with throttling valves, as a back-up to the gasinjection system. Use of the throttling valves was seen as having the disadvantage of requiring alarge pressure drop to be effective. However, installation of ESPs on Forties Echo has meantthat pressure is available to overcome the throttling valve pressure loss, and this combined withthe fact that gas is often not available when FE is started up has meant that use of the throttlingvalve on the 12” riser has become the standard means of avoiding severe slugging at start-up.As water cut has risen, surging at start-up has become more pronounced such that even withthe throttling valve shut-in to its minimum stop (25% open) the system was regularly tripping outon high separator level at start-up.

Forties have recently reduced the minimum stop position to 10% and this has enabled opera-tions to bring FE on line with minimal surging and without high level separator trips. Datacollected during the November 1993 field trials illustrated the relatively steady nature of a start-up of the 12” line when shutting in the throttling valve to its minimum stop condition (9).

3.4.3 Slugging Produced by Transient Effects

In addition to the mechanisms already discussed, slugs may be produced as a result of tran-sient effects such as pressure or flowrate changes. For example, if a line operating in stratifiedflow is subject to an increase in gas flowrate, or total production rate, one or more slugs maybe produced as the equilibrium liquid level drops towards a new steady state condition.

BP has access to two general purpose transient multiphase programs. These are PLAC, devel-oped at Harwell as part of a Joint Industry Project, and TRANFLO, developed at Sunbury andnow available through Harwell. OLGA is another commercially available transient simulator.Current thoughts on the relative merits and applicability of these codes, together with a detaileddiscussion of transient multiphase modelling is given in Section 7 of this manual.


Section 3. Oil/ Gas Pipeline Design BP Multiphase Design Manual

3.4.4 The Effect of Slugging Flows on Process Plant

(a) Control Philosophies

It has already been mentioned that the intermittent nature of slug flow results in the separationplant experiencing variation in liquid and gas flow. A separator, or slugcatcher subject to slug-ging conditions will thus experience variations in:

l Liquid level

l Gas pressure

l Liquid and gas outflow.

The exact nature of these variations will depend on the way in which separator/slugcatchercontrols are set up.

In some cases the first stage separator is set up on level control. If the control system is set torespond rapidly to changes in level, the liquid level may rise only slightly, but the liquid outflowwill increase sharply. Under these circumstances the liquid disposal system (valves andpumps), and the downstream liquid handling plant will have to deal with rapid variations in liquidflowrate. The problem with this type of system is that the disposal pumps and downstreamliquid processing plant need to be sized for flowrates well in excess of nominal maximumproduction rate. In addition, the separator liquid residence time is reduced during periods ofslug production so that oil/water separation efficiency may deteriorate.

The usual way in which slugs are catered for is through use of a large separator, or slugcatcher,where the increase in liquid flowrate during slug production is handled by allowing the liquidlevel to rise. Provided the liquid level remains within set limits the liquid disposal rate may bekept constant. The vessel is then said to be on flow control. Once the liquid level rises above acertain point the liquid flow controller has to be reset to a higher value to avoid high level shut-downs. The advantage of this type of system is that the flowrate of liquid to the downstreamplant is maintained near constant. The disadvantage is the large size of vessel required toaccommodate the slugs.

The two types of control philosophy discussed, level control and flow control, represent the twoextreme methods of operation. In practice some blend of the two may provide the solution forany particular application. In order to optimise the liquid level and disposal control systems adynamic model of the plant should be developed during design. Dynamic modelling is furtherdiscussed below.

In addition to handling the variations in liquid flowrate during slugging, the reception facilities willhave to be designed to cater for the large variation in gas flowrate during slugging.

In fact generally, it is the gas plant which is more susceptible to operational upsets during slug-ging than the liquid plant. As with control of the liquid level, there are two philosophies forhandling gas production from the separator or slugcatcher.

(al) Fixed pressure control

In this system the excess gas flowrate occurring during production of the gas bubble passesrapidly through the separator to the downstream gas plant.


BP Multiphase Design Manual Sectton 3. Oil / Gas Pipeline Design

(a2) Allow Pressure in Separator to Rise

Here at least some of the excess gas production is held back in the separator as the pressureis allowed to rise. This type of system will impart a smaller perturbation on the downstream gasplant than one with fixed pressure control.

In this case the separator is acting as a gas accumulator. Clearly this type of system requiresthe normal operating pressure of the separator to be set at some level below the design pres-sure.

(b) Dynamic Simulation of Process Plant

In order to accurately assess the effect of slugging on the process plant it is necessary to takethe output from a slugging model and input this into a dynamic model of the process plant.BPX have used the dynamic simulation package SPEED-UP on a number of recent develop-ments studies including Bruce WAD, Cyrus-Andrew, and Cusiana. This program is used anddeveloped by the Process Simulation Group of ESS Sunbury. SPEED-UP can be used tomodel a wide range of items or process plant including pumps, compressors, valves, and alltheir associated control systems.

A slugging model, such as NORMSL, see Section 3.4.1, produces output in terms of gas andliquid production rates with time. In order to accurately assess the interaction between theprocess plant and the slugging pipeline, NORMSL has been interfaced with SPEED-UP. In thisway any pressure changes in the separator or slugcatcher, caused by slugging, are fed backdirectly into the pipeline model where they may affect the production rate of subsequent slugsand bubbles.

The dynamic model can be used to assess the ability of a preliminary design to handle sluggingconditions. If the model suggests that slugging flow will create unacceptable plant conditions itmay be possible to produce an operable system by modifying or changing the control system.On other occasions new or larger vessels may be required.

The Process Simulation Group use the combined NORMSUSPEED-UP model to assess therelative merits of various active slug control schemes, such as the use of separator inlet throt-tling. They are also conducting field trials to validate the findings of such modelling activities.

The PLOT team of BPX Aberdeen use a dynamic simulation package to assist in optimisingplatform topsides plant conditions and assessing the effect on the process plant of changes infeed rate and operating conditions. PLOT have recently incorporated the BPX slugging modelNORMSL into their dynamic model of the Bruce topsides to enable them to assess the effect ofslugging from the Bruce Western area on the Bruce process plant.




LEFT BLANK



References

1. “Two-Phase Flow in Pipes”Course Notes by J.P. Brill and H.D. Beggs

2. BPE File Note PSC/2/808/008“Recommended Pressure Loss and Hold-up correlations for Crude Oil/Gas andGas/Condensate Flowline Systems”(P. Sugarman, 9-l -90).

3. “Wytch Farm Infield Flowlines Simulation”BPE File Note PSC/W8/75-1(P. Sugarman, 1 l-4-90).

4. BPE File Note PSC/2/808/008“Two-Phase Downflow Riser Correlations”(J.T.C. Hau, 23-l O-90).

5. The Multiphase flow Database‘Comparison of Multiphase Correlations with Pipeline Data”Hat-well (23-3-91).

6. BPE Record Note ME/N40/10“Maximum Erosional Velocities in Duplex Stainless Steel Production Flowlines and ManifoldPipework and Carbon Steel Water Injection Flowlines”(R.A. Homer, 22-10-86).

7.

8.

BPR File POB/l51“New Model Flowloop with Wax Deposition Cells”(P.C. Anderson, 10-5-91).

BPE Report PSC/2/808/008“Summary of Field Data on slugging Multiphase Systems - BP and Other Data”(P. Sugarman, 21-6-91).

9. XFE Report XFEIOl l/94“Performance of the Forties Echo-Alpha 12 Inch Production Line With Gas Injection and RiserTop Choking”(T.J. Hill, P Sugarman, 17-2-94)

10.

11:

BP Report ESR.94.ER.076“Erosion Guidleines”(Materials and Inspection Engineering Group, 22-7-94).

BP Report ESR.94.ER.016“A Corrosion Philosophy for the Transport of Wet Hydrocarbon Gas Containing CC,”(Facilities Engineering, BPX / Materials and Inspection Group, ESR , 28-8-94).


Section 3. Oil/ Gas Pipeline Destgn BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 3. Od / Gas Pipeline Des&n

3A Slug Flow Design Method

3A1 Introduction

As described in the main body of the design guidelines (Section 3.4.1 (a)) the pioneering fieldwork on steady-state slug flow was carried out by Professor Jim Brill (Tulsa University) atPrudhoe Bay in 1978 and 1980 (1). The resulting correlation for slug length was derived usingstatistical regression techniques on gas and liquid flowrate, and diameter.

In the mid 1980’s BP began their own programme of field data collection on lines in slug flow.The ‘Field Data’ section of the bibliography of BP Research reports on multiphase flow lists therelevant titles. It quickly became evident that the original Brill equation (and the modificationdue to Norris) was very conservative in predicting both mean and maximum slug lengths over avariety of flowrates and diameters. This provided the incentive to develop an improved methodthat sought to take into account a greater number of the relevant variables than the original Brillmethod, and to do so using a mechanistically based approach.

BP Exploration have been involved in the development and implementation of this method,which was installed into the MULTIFLO computer programme early in 1990, thereby making itgenerally available within the BP group. In addition, a number of papers, covering test rig andfield data collection and the development of part of the slug flow method, have been publishedby BP in the open literature.

3A2 Key Variables in the Prediction of Slug Characteristics

As stated in Section 3.4.1 (a) the key parameters required for the assessment of the perfor-mance of separation and associated downstream facilities are the average slug volume andfrequency, and the greatest likely slug volume and its frequency of occurrence. To assess thedesign of pipework supports the combination of slug velocity, frequency and liquid holdup isimportant.

These parameters are all influenced, to a greater or lesser extent, by the following variables:

VW = gas superficial velocityV,, = liquid superficial velocityD = pipeline diameterL = pipeline length6 = pipeline inclinationp, = gas densityp, = liquid densitypg = gas viscosityt.~, = liquid viscosityU, = surface tension

The aim of the work described here has been to develop a mechanistically-base method forslug characterisation that reflects the contributions of many of these different variables. Thisgives much greater confidence in the method output over that of a straight correlation based onstatistical regression of limited using only one or two variables.

-



3~4.3 Development of the Average Slug Frequency Method

The physical process causing normal slug flow in (near) horizontal pipe flow is the formation andgrowth of a wave on the surface of a stratified liquid film. This wave results from a flow of gasover the liquid film. The wave may grow until it reaches the top of the pipe at which point thegas flow is sealed behind the incipient slug. As the gas velocity was much greater than thewave, there is then an acceleration of the wave/incipient slug up to a velocity approaching thatof the bulk gas. This increased wave/slug velocity results in liquid in the film ahead of thewave/slug being picked up and incorporated into the slug body. Liquid is shed off the back ofthe newly formed slug, but at a slower rate than the pickup rate. The slug keeps growing in thisway until it reaches the end of the liquid film shed by the slug in front (with the film shed from aslug being less deep than the equilibrium stratified liquid film). A stable length is then main-tained.

The fundamental parameters determining the average slug length are therefore the slug forma-tion frequency, the depth of the equilibrium stratified liquid film ahead of the slug, the behaviourof the liquid film shed from the rear of the slug, and the liquid holdup in the slug body.

Of these four parameters the one with the least reliable prediction method, before the BP R&Dprogramme, was the slug frequency. Correlations based on small diameter and/or a limitedrange of fluid properties have been available for some time, but none performs particularly wellagainst the field data collected by BP since 1985.

What has now been developed is a relationship between a dimensionless slug frequency andthe equilibrium stratified liquid holdup (HJ that would result from the prevailing gas and liquidflowrates and physical properties if slugs were not being produced (2). The exact nature of therelationship has not yet been derived mechanistically, and so a correlating line is used whilstfurther work on the mathematics is being undertaken. However, this is a big step further onthan the simple regression techniques, which did not aim to describe the physics behind thephenomena.

The dimensionless frequency has the form:

where:

F, = the actual slug frequencyV, = in-situ gas velocityV, = in-situ liquid velocity

H,,, V, and V, are calculated from the equations (due to Taitel and Dukler (3)) for determiningthe depth of the equilibrium stratified liquid film given the gas and liquid flowrates, pipelinegeometry, and fluid properties.

Gas phase

&I w s, -Ti Si -A, p, g sin6


-


Liquid phase

dP

A, -I 1 = -7w, S, +T~ Si -A, p, g sine

dx I

where:

A,=AH,A, =A(1 -H,JS, = gas wall contact lengthS, = liquid wall contact length

h = gas shear stress to the pipe wall7wt = liquid shear stress to the pipe wall7i = inter-facial shear stress between gas and liquid

At equilibrium

The two equations may then be solved for H,,.

The form of the slug frequency method that correlates the dimensionless frequency against theequilibrium stratified liquid holdup, is given below. This method was published in (7):

FS D[ 1 ’= -24.729 + OJJo766 eP.91209 l HW + 24.721 e(O.20524 ‘HW

Yn

where:

F, D[ 1 ’ F, D~ = - (1 - 0.05 V,J . Do.3

vrl7 vm

D in m

V,, VW and V,, in mfs0.068H,’ = H,, 1 -[ 1Vsl

This relationship covers both published and in-house data from test rigs and producing oilfields.The diameter range is from 1.5 to 24 inches, with fluids such as air and water, nitrogen anddiesel, and actual produced reservoir fluids involved in the studies.

The inter-facial friction factor, fi, used in the development of the correlation, was that due toAndritsos and Hanratty i.e.

For V, 5 1.5 m/s



fifi = = fgfg

For V, 2 1.5 m/s

fifi = = f,f, [l[l + + (vsg1.5(vsg1.5 -1) 0.751 -1) 0.751

Further work is underway to revise the slug frequency method by the inclusion of additional testrig and field data, and by reviewing the selection of the inter-facial friction method.

3A.4 Slug Holdup Method

Currently there is one widely recognised method for the prediction of liquid holdup in slugs, dueto Gregory (4). This is presented as a function of mixture velocity, shown below:

11

HISHIS = = 1 (with Vr,, in ft/s)1 + 1 + (VJ28.4)'-(VJ28.4)'-

This correlation gives reasonable performance with two-phase mixtures (e.g. air and water, ordry crude and associated gas). However, when there is water in the crude the slug liquidholdup tends to be much higher than predicted.

This does not affect the calculations of average slug volume (as the slug model is a massbalance determined by the slug frequency, which is calculated independently of slug holdup).However, calculations of slug length must always be considered in conjunction with thepredicted slug liquid holdup.

The slug holdup becomes an important parameter when considering pipework loads on bendsand supports (see relevant section of the Design Guidelines).

3A.5 Use of the Creare Slug Model

Models of slug flow are essentially mass balances of gas and liquid flow through theslug/bubble unit. The main difficulty is encountered in predicting the behaviour of liquid beingshed from the back of the slug.

The principal workers in this area have been Dukler and Hubbard (5) who wrote a model forslug flow in 1975. The model requires slug frequency and slug holdup as inputs. The modelwas revised by Crowley of Creare inc. during 1986 to improve the solution algorithms (6).

The solution process involves first calculating the length of the gas bubble/liquid film section.Three variables are involved, and they are interdependent (therefore requiring solution of twosimultaneous equations) - the gas bubble length (LJ, the minimum film velocity (V,J and theminimum film holdup (H,&. The gas bubble length is calculated as a function of the minimumfilm velocity. The detailed equations for all inclinations are available from BP Research. Forhorizontal flow the solution sequence is given below:

Vti = velocity of slug front



L, = mean length of slugs

Slug frequency (FJ and holdup (HJ are obtained from the methods described in the precedingsections.

v,, = 1.3 v, (with V, in m/s)

e “mL,=- - -

L I1

wn “anin

where:

a = 0.046

I- h&-u 1‘cmin = ‘sf -

L bm, 1

From the above equations Lb can be expressed as a function of H,,. The second independentequation also expresses the gas bubble length as a function of the minimum liquid film fraction:

“sf- H,, + 0.3 (H,, - H,,,)

F, 1.3 (HI, - H,iJ 1The two equations are then solved for Lo.

Finally the mean slug length is obtained from:

“sL, = I 1- ‘L,

Fs

This is the output of slug length in MULTIFLO.

3~6 Development of the Slug Length Distribution Model

Since the work by Brill there has been an assumption that the distribution of slug lengthsusually follows a log-normal distribution. The large amount of test rig and field data collected byBP indicates that this is not valid as a general assumption. The log-normal case tends to bethe extreme form of distribution applicable - extreme in the sense that the ratio of 0.001 proba-



bility slug length to mean slug length is higher than other distributions that have since beenconsidered for slug flow. The implication of this is the possibility of over-sizingslugcatcher/separator vessels.

Other distributions that are applicable under varying circumstances are normal, lambda andinverse Gaussian. Each distribution method requires the determination or assumption of one ormore parameters, in addition to the predicted mean value.

Currently the lambda distribution with a power of 0.8 is used. This gives a 0.001 to mean ratioof 2.6, as opposed to the ratio of 4.12 for the log-normal distribution (with the assumption of astandard deviation of 0.5).

Work is currently underway on the prediction of which distribution (and the values of parame-ters required to apply that distribution) is best applied under any given set of conditions.

3A.7 Development of the Maximum Possible Slug Length Method

Under some conditions it may be possible to have only one slug in a line. This could result fromstratified flow that could easily be disturbed by, say, a slugging well to form a flowline slug, orfrom very low frequency slug flow, or a relatively short pipeline. Having only one slug in a line isa special case that requires some additional analysis.

The mean slug length prediction described above assumes that each slug grows until it reachesthe film shed from the slug ahead. If the time period between slug generation is significantlylonger than the residence time of a slug in the line then the slug may not reach the film shed bythe slug ahead of it. The slug length will then be dependent on the additional factor of pipelinelength.

The procedure for calculating the ‘maximum possible slug length’ begins with the calculation ofequilibrium stratified liquid holdup (HJ, as described earlier. The equations below are thenused to calculate the maximum length.

Vs = velocity of liquid in slug bodyV, = velocity of slug rearV, = velocity of liquid in equilibrium stratified liquid filmL = pipeline lengthLmax = length of maximum possible slug

vs HIS - VII Hlev, =

HIS - HieV, = 1.2 V, + 0.35 (g D)“.5T = L/V,Lmp( = cVsf - V,J . -I-

This method is sensitive to the value of slug liquid holdup, but can be used as a good estimateof the maximum possible slug length that could be obtained in a situation with only one slug ina line. It is not currently available in MULTIFLO, but can be accessed via the Multiphase FlowGroup, BP Exploration, Sunbury.


BP Multiphase Design Manual Section 3. Uil/ Gas Pipeline Design

3A8 Implementation and Use of the Slug Flow Design Method

There are two ways of using the BP slug flow design method. The first of these is viaMULTIFLO, which is developed and supported by the Multiphase Flow Group and runs on aVAX computer. This group may be contacted to carry out design studies, or your local VAXcluster may be used to run the programme.

Contact: Phil Sugarman (MSMail Sugarman,Phil or phone +44 1932 762882)

The alternative is to run the model on a Hewlett-Packard workstation. This version contains alarger number of options for calculating slug holdup, frequency and length than are at presentinstalled in MULTIFLO.

Contact: Trevor Hill (MSMail Hill, Trevor or phone +44 1932 763298)

3k9 Anticipated Future Developments

1. The formulation of an improved method for slug liquid holdup prediction that takes intoaccount water cut as well as gas and liquid superficial velocities and line diameter.

2. The broadening of the slug frequency method to cover inclined lines, and high viscosityliquids.

3. The completion of the slug length distribution model that will include the effect of gas andliquid flowrate on the distributions.

Principal References

1. Analysis of Two-Phase Tests in Large Diameter Flow Lines in Prudhoe Bay Field,Brill, J P et al, SPEJ June 1981,363-378

2. Further Work on a Method for Predicting Slug Frequency in Multiphase Flow,Hill, T J, RCS Branch Report 123 959, 3.9.90.

3. A Model for Predicting Flow Regime Transitions in Horizontal and Near-Horizontal Gas-LiquidFlow,Taitel, Y, and Dukler, A E, AIChEJ, 22,47-55, 1976.

4. Correlation of the Liquid Volume Fraction in the Slug for Horizontal Gas-Liquid Slug Flow,Gregory, G A, et al, Int J Multiphase Flow, 4, 33-39, 1978.

5. A Model for Gas-Liquid Slug Flow in Horizontal and Near Horizontal Tubes,Dukler, A E, and Hubbard, M G, Ind Eng Chem Fundam, 14, 4,337-346, 1975.

6. State of the Art Report on Multiphase Methods for Oil and Gas Pipelines,Crowley, C J, Creare Report TN-499 to AGA, 1986.

7. Slug Flow - Occurrence, Consequences and Prediction,Hill, T J, and Wood, D G, SPE 27960, 1994.




LEFT BLANK

--

-



3B Slug Generation Dips

3B.l Introduction

A significant effort has been made by BP, and by other oil companies and research establish-ments, to understand and characterise steady-state slug flow. Such a flow occurs with steadyinput gas and liquid flowrates and in pipelines that are normally assumed to be horizontal. Inthese cases slugs are formed from hydrodynamic waves, caused by a gas flow over a stratifiedliquid film, which rise to bridge the pipe. This work is covered in Appendix 3A.

Even for non-horizontal lines, or lines that change inclination, there is available to BP a qualita-tive assessment of the difference in behaviour from a straight horizontal pipe, at flowrates thatwould produce slug flow in the horizontal pipe.

A second classification of slug formation is required for systems with a low liquid loading and anuneven pipeline topography. If the pipeline were horizontal then the liquid flowrate would notnormally be sufficient for slug flow to occur. However, if there is a dip at any point in the line itmay be possible for liquid to accumulate until there is a sufficient quantity to produce a slug outof the end of the upward sloping line leading away from the dip.

This appendix addresses that possibility. It is based on work done largely for the Miller gasexport line. This line normally runs in dense phase flow. However, in a depressurisation situa-tion liquid will drop out. A prediction of possible slug sizes arriving at the downstream end ofthe line was required to enable an assessment of slugcatcher size to be made.

3B.2 Description and Theory

At low liquid flowrates, liquid transport out of a dip will definitely occur at a gas rate sufficient toproduce annular-mist flow. This appendix concentrates on the liquid removal that occurs atmuch lower gas rates than the annular-mist transition when the dip angle is relatively shallow.The liquid may be removed as slugs or as a film.

The study began with experimentation. A volume of liquid was put into a shallow angle dip,with no further liquid inflow into the dip, and with no gas flow. As the gas flow was increasedthe liquid in the dip was disturbed - initially to form waves, and subsequently to form slugs ifsufficient liquid was in the dip at the start. As the gas rate was increased further the liquid wasproduced out of the upward sloping line leading out of the dip, in the form of a film, slugs ordrops, or a combination of these.

A number of quantities are defined in the analysis:

Critical gas velocityThat gas velocity above which all liquid is produced from the dip.

maximum stable liquid accumulationThe maximum volume of liquid that can be retained in a pipeline dip at a given gas superficialvelocity (i.e. no liquid production out of the end of the upward sloping pipeline segment).

Maximum random slug sizeThe largest liquid slug volume likely to be produced from a dip under steady-state operation.



(a) Critical Gas Velocity

This variable is calculated from the momentum balance equations for an equilibrium stratifiedfilm (due to Taitel and Dukler).

Gas phase

- A, P, 9 sW3

Liquid phase

dPA, -[ 1 = -T,, S, + 7i Si - A, p, g sin(e)

dx I

At equilibrium

The equations may be re-arranged to give:

Ll Sl 1 11 1i 1 k4k4 %I%I~~ = = TiSiTiSi -+--+- + + ~ - (P, - PJ 9 sin@)

AI A, Ag Ag

For given fluid properties and pipe characteristics, the right hand side of the above equation isdependent on only the gas velocity and film depth.

The solution requirement is a gas velocity (V,,,J such that, at any liquid film depth in theupward sloping line downstream of the dip, the film flow direction is co-current with the gas.For the 2 inch air-water test rig the predicted value of Vgscrit corresponded closely to the obser-vations.

One of the key parameters required during the calculation procedure is the gas-liquid inter-facialfriction factor. This variable has an element of uncertainty in its prediction. Numerous efforts ata better definition have been, and are being, made.

(b) hhxhum stable liquid accumulation

The magnitude of this quantity is dependent on the prevailing gas velocity, as well as the geom-etry of the pipeline. The calculation procedure is based round the growth and decay of slugs inthe uphill section of line. The largest amount of liquid contained in a dip occurs when the uphillsection is in slug flow, but with the slugs decaying just before the end of the section.

-



If the theoretical average slug decay rate over the uphill section length is equal to, or greaterthan, the average slug growth rate over the same section, then any slugs formed will decay bythe section end. The total volume of liquid in the section will pass through each slug in thislimiting case. Therefore, assuming that the liquid holdup in the slug is constant along the line,and that the slug rear velocity may be approximated by a steady-state correlation, themaximum liquid accumulation (C&J may be calculated as:

(0.16 V, + 0.59)Qeqm = 1 A HIS L,

(1.16 V,,, + 0.59)

H,, may be evaluated using a steady-state correlation based on gas and liquid superficial veloci-ties, A and LP are the pipe section cross-sectional area and length of the uphill section respec-tively. This equation is valid for gas superficial velocities less than VWti,.

A rapid increase in gas velocity, or the passage of a pig through the dip, have the potential tosweep out the bulk of this maximum liquid accumulation.

(c) Slug frequency

Unlike the case described above, real flowlines generally have a liquid inflow into the dip froman upstream source. In this situation a balance will be set up in which liquid will be producedfrom the end of the uphill section of pipeline at the same rate (on average) as the liquid inflowinto the dip. However, the liquid will not always be produced steadily. Sometimes it will beproduced intermittently in the form of slugs. If this is the case it is necessary to calculate theslug formation frequency in order to be able to predict the volume of the slugs leaving the uphillsection.

Slugs are formed at or just after the point of minimum pipeline elevation from the liquid filmcollected there as a result of both liquid inflow from the upstream source and the return of liquidshed from preceding slugs. Calculation of the slug frequency is dependent on these two liquidflowrates, the gas flowrate, and the depth of liquid at the pipeline low point required beforeslugs will form.

The assumptions currently made are that the liquid inflow from upstream is constant, and thatthe surface of the liquid accumulation in the dip is horizontal. However, it will be possible todevelop the method to take into account cases when these assumptions do not hold.

The liquid flowrate draining back into the dip from a previously formed slug is calculated from amass and momentum balance on the liquid film shed from the back of a slug. This flowrate iszero for a short period after slug formation, and then increases to a steady value as thedistance of the back of the slug from the dip increases.

Slug formation will, on average, occur when the depth of liquid in the low point of the dipexceeds a certain critical value (dependent on the gas flowrate). This critical depth is calculatedusing the procedure developed by Taitel and Dukler for predicting the onset of slug flow from astratified film.

Slug frequency is then determined by integrating the flowrate into the dip over time, convertingthat volume to a liquid depth, and stopping the procedure when the critical liquid depth isreached. The slug formation frequency is the inverse of the time required to reach the criticaldepth.


Section 3. Oil / Gas Ptpellne Design BP Multiphase Design Manual

(d) Slug lengths leaving uphill section

The volume of slugs persisting to, and leaving the end of, the uphill section will, on average,equal the volume of liquid entering the dip over the time period between the formation of slugs.

It is, however, important to note that a real system has a degree of fluctuation about theseaverage values. The length of slugs leaving the dip will be distributed about the mean value. Asuccession of short slugs would mean an increasing excess of liquid over the maximum stableliquid accumulation. A slug much longer than the average could then result, which decreasesthe liquid inventory in the system back down to the maximum stable accumulation.

A factor of 10 times the average is recommended as the safety margin in estimating the largestslug likely to be produced at the prevailing gas and liquid flowrates.

(e) Complicated topographies

The above discussion has centred on a dip with a steady inflow of liquid. In a real line it ispossible that there will be several hills and dips, with unsteady liquid production from one dip tothe next. The methods described above may be used to track the liquid movement through thesystem. However, the dependence of the output on the variation in slug length must beemphasised.

3B.3 Availability of Methods

The implementation of this method currently is a programme written for Hewlett-Packard series200/300 computers running BASIC 5.0, available via Team 121 (Multiphase Flow) at the BPExploration, Sunbury.

Principal Reference

A Mechanistic Model for the Prediction of Slug Length in Gas/Condensate Lines,Wood, D G, RCS Branch Report 123 895 dated 255.90.



3c Severe Slugging

3C.l Introduction

The possibility of significant surging occurring at low flowrates through a pipeline ending in ariser was first identified in the early 1980’s. Work at Tulsa University by Schmidt, and in the fieldby Yocum, had shown that if the pipeline to the base of the riser was inclined downwards, andif the flowrates were low enough, then there could be a liquid accumulation at the base of theriser that would result in the severe slugging cycle described in the main body of the text.

At about the same time BP were planning the development of the SE Forties reservoir, byrunning flowlines from above the reservoir (originally from subsea templates, and finally from aminimum facilities platform) back to the existing Forties Alpha platform. These schemes involvedflowlines sloping downward (about 0.1”) to the base of a 120 metre high riser. Concern wasexpressed about the possibility of riser induced severe slugging, which led to experimental andtheoretical investigations by BP Engineering and BP Research. The results of these investiga-tions, and subsequent experience with the SE Forties flowlines, provide the basis for thisappendix on severe slugging. The detailed reports covering the work are listed under ‘SevereSlugging’ in the bibliography of BP multiphase flow reports.

3C.2 Description of Severe Slugging

The four basic stages of severe slugging have already been described in the text. They are:

l Slug Generation

l Slug Production

l Bubble Penetration

l Gas Blowdown

A 50 mm diameter pipeline-riser system was constructed at the BP Sunbury site to investigatesevere slugging. The pipeline section was 50 metres long and the riser 15 metres high. Avideo of this test rig, showing the different multiphase flow regimes, including severe slugging, isavailable, entitled Slugs in the Pipeline’.

Assuming that bridging (which occurs at the start of the slug generation stage of the cycle,when liquid completely blocks the bend to the pipeline-riser) has occurred it is relativelystraightforward to describe the stability of the bridge (and hence whether severe slugging willoccur at the given flowrates and pipeline inclination and length), and also the cycle time ofstages (1) to (3). Characterisation of the fallback (during gas blowdown), and the prediction ofthe volume of liquid left at the base of the riser immediately after fallback and bridging, is moredifficult, thus introducing a greater degree of uncertainty into the calculations of overall cycletime. The calculation procedure to determine the occurrence of severe slugging is describedbelow.

The equation that describes whether or not severe slugging can occur at the given combinationof gas and liquid flowrates, pipeline diameter, inclination and length, is derived as indicatedbelow, and involves the following variables:



L = pipeline length (m)L, = length of liquid back-up from riser base (m)D = pipeline diameter (m)A = pipe cross-sectional area (maH = riser height (m)I3 = angle of riser to horizontal (degrees)h = length of liquid column in riser (m)p, = gas density at separator pressure (kg/m31p, = liquid density (kg/m31V, = superficial gas velocity at separator pressure (m/s)Vd = superficial liquid velocity (m/s)g = acceleration due to gravity (m/s3M = gas molecular weight (kg/kg mol)n = number of kg moles of gas in pipelineI? = universal gas constant (J/K/kg mol)P = pipeline pressure (N/m*)P = separator pressure (N/m*)T%I pipeline temperature (K)H, = liquid hold-up in pipelineV, = gas volume in pipeline (m3)

The gas and liquid mass flowrates into the start of the flowline are assumed to be constant. Atthe boundary of the severe slugging regime the liquid inflow to the start of the line exactlybalances the rate at which liquid fills the riser after bridging (i.e. L, = 0). The second conditionfor determination of the boundary is that the rate of gas pressure increase in the pipelineupstream of the bridge is produced exactly by the rate of gas inflow. The rate of gas pressureincrease is equal to the rate of liquid head increase in the riser. The value of liquid hold-up inthe pipeline is calculated from open channel flow equations, as the in-situ gas velocity is verylow during this stage of the severe slugging cycle.

dP dh

- = b, -P&l - sinI (1)dt dt

dP d(nM) RT- = (2)dt dt M “Ll

vg = L (1 - HJ A (3)

d (n Ml= “sg P, A (4)

dt

dh- = v,,

dt(5)

Substituting (3) and (4) into (2) and then (2) and (5) into (1) and rearranging, gives the equa-tion quoted in the text, given that:


-


P, R TP = and sin f3 = 1

=P

M

“w L (1 - H,) (P, - PJ 9- =

“,I P=P

(6)

If this equation shows LHS 5 RHS then severe slugging will occur if the pipeline slopes down-wards to the riser, and if the flow in the pipeline is predicted to be stratified at the prevailing in-situ superficial gas and liquid velocities.

This compared well with the observed boundary on the test rig, and also gave good predictionsof cycle times (stages (1) to (2)) when equation (1) was integrated over the height of the riser.

The bubble penetration stage may also be modelled simply. If LHS < RHS in equation (6) thenthere will be liquid backed up the flowline (i.e. L, > 0), as well as a full riser (h = H). Once theliquid column in the riser has reached the top of the riser (maximum liquid head at the base ofthe riser) the gas inflow at the start of the pipeline is no longer required to boost the pipeline gaspressure. The gas flow therefore begins to push the accumulated liquid along the flowline at avelocity that may be calculated from the inlet gas mass flowrate, the maximum pipeline pres-sure, and the liquid film hold-up.

dL PseP

1- = V

sgdt (Psep + pI g HI (1 - HJ

The rate of liquid production from the top of the riser at this stage is simply:

d Ls= v,, + - (1 - $1

dt

Once the bubble front reaches the base of the riser there begins a period in which the liquidcolumn in the riser is decreased in length, and is accelerated by the ever-increasing imbalancebetween the upstream gas pressure (which remains relatively high) and the decreasing liquidhead. The acceleration of the liquid column in the riser is given by:

g W -h)=

h

which obviously approaches infinity as the bottom of the liquid column reaches the top of theriser. Therefore there is a requirement to model downstream pipework and restrictions, whichdetermine the upper limit of the liquid production rate.

An empirical approach was used during the test rig studies to estimate the liquid left in the riser,



after bubble penetration and gas blowdown, at the time of bridging. This liquid would then fallback to the bottom of the riser to produce a liquid column of a height that is used as thestarting point in the integration of:

dh

dt

to obtain a cycle time.

Obviously the field examples are more complicated - the gas may not obey the ideal gas law,there may be gas transfer into and out of solution in the liquid with pressure change, thepipeline might not be of a constant inclination, and the length of the line might be such thatthere is a variation of pressure and liquid hold-up along the line.

The simulator used to make calculations for SE Forties, whilst being founded on the physicalprinciples described above, also seeks to take into account some of the complications notpresent on the test rig. However, equation (6) still provides a rule of thumb for part of thesevere slugging boundary for more complex systems.

3c.3 Test Rig Studies

The test rig described above was used to develop a flow regime map for the pipeline-riser,flowing air and water with the separator at atmospheric pressure. Good agreement with thetheoretical boundary, and with the theoretical cycle times was found. This gave the necessaryconfidence in our basic understanding of the phenomenon to enable some of the more detailedrequirements of the field case to then be worked on in the VAX-based simulator.

Measurements were made of the liquid flowrate into the separator during the bubble penetra-tion stage. As described above, the mathematical models need to have some pipework orrestriction downstream of the riser-top to make the simulations realistic otherwise there wouldbe an infinite liquid production rate as the last portion of liquid was produced from the riser.

The test rig was also used to investigate the flow behaviour at low flowrates in systems with ahorizontal line and an upward sloping line upstream of the riser-base. Both of these cases alsoproduced significant surging, but not the classic severe slugging in which liquid completelyblocks the pipeline-riser bend. In these cases there was a liquid accumulation, but with gasbubbling through at all times. This observation is significant, however, especially for deep watersystems (e.g. Gulf of Mexico) where there is the possibility of a significant liquid head build-up inthe riser at low flowrates even if the pipeline slopes upward to the base of the riser.

3c.4 Ways to Eliminate Severe Slugging

There are a number of ways to reduce the extent of, or completely eliminate, severe sluggingresulting from low flowrates in a pipeline-riser system.

(a) Pipeline Orientation

The first is to ensure that the pipeline route slopes up to the base of the riser, at least over adistance immediately upstream of the riser that is several times the riser height. This may mean


BP Multiphase Design Manual Section 3. Oil/ Gas Pipeline Design

a slightly longer pipeline route than a direct line, but the extent of the surging at low flowrateswill be reduced, as full severe slugging cannot occur with an upward sloping line to the riserbase.

(h) Gas Injection

The second technique is to have the facility to inject extra gas into the base of the riser. Gasinjection into the base of the riser continually lifts liquid out of the riser, preventing the build-upof liquid and subsequent seal to the gas flow.

Test rig studies were carried out on riser-base gas injection after the work described above onthe severe slugging cycle. The principle was demonstrated successfully. Even gas injectionrates that are not sufficient to lift liquid continually out of the riser are of benefit as the gasreduces the maximum head achievable in the riser, therefore reducing the cycle time andseverity of the severe slugging cycle. The amount of gas injection needed to prevent severeslugging is such that the combined injected and produced gas flowrate up the riser must putthe riser into the churn flow regime at the given liquid production rate.

Studies were also carried out to optimise the design of the gas injection tee, to minimiseerosion. The final design comprised a jacket around the riser into which the injection gas wasdirected. The gas then flowed into the riser through eight holes spaced equally around theriser. The holes were at the bottom of the jacketed length of riser.

Thirdly, gas injection into the start of the pipeline may be used to move the pipeline out of thesevere slugging regime, by increasing the ratio of Vsg to Vsl above the critical value defined inequation (6). This either requires surplus gas at the pipeline start, or the addition of a gas lineout from the pipeline end.

(c) Choking

A fourth method is to install a control valve on the top of the riser. On shutting this valve a posi-tion is reached whereby the frictional pressure drop across the valve acts to stabilise the gas-liquid flow up the riser. Any acceleration of liquid up the riser due to a decrease in liquid head inthe riser (caused, say, by a gas bubble entering the base of the riser), is counteracted by theincrease in frictional pressure drop across the valve as the liquid accelerates. The penalty ofthis way of eliminating severe slugging is that the pressure drop across the valve will be of theorder of a riser height of liquid, thus imposing a significant extra back-pressure on the system atall times.

A secondary effect of the valve-induced back-pressure, as well as stabilising the riser, may bethat the flow regime in the pipeline section is moved out of stratified flow towards bubble flow.This latter effect, if it was predicted to be sufficient to prevent severe slugging, could also beachieved solely by raising the separator pressure.




LEFT BLANK



Slug Flow in Hilly Terrain

Figure 5 shows a method for predicting slug characteristics in hilly terrain system

I *“q>soak? I

Figure 5. Hilly Terrain Slug Sizing Procedure--



The following comments relate to this method, pointing out the sources of information andsome of the areas in which further work is required:

1. L, mean, 1, mean, kl 000 and L,,lOOO determined from MULTIFLO (in which the equilibriumholdup (H,J method uses Blasias for gas and liquid friction factors, Fi = F, for inter-facial frictionfactor, and (Vg - V,) for inter-facial velocity).

2. The frequency correlation which results in the the above lengths was derived for near horizontalonly. Test rig studies, performed recently, show that above about 7 m/s Vsg the inclination (upto 2 degrees) has little effect, so not conservative at these velocities. However, for lower V,‘s,down to 5 m/s, but same GLR, the rig is showing slug frequency not very dependent on V,,,staying roughly at the 7 m/s V, (and associated V,,) value. For horizontal lines slug frequency ishighly proportional to V,,. Application of the horizontal method to upwardly inclined lines will beconservative.

i.e. if slug frequencies in upward lines are higher than horizontal at the same flowrates, whichthe test rig shows, then our horizontal slug lengths used in the attached logic tree will beconservative. Improvements to the prediction of slug frequency in inclined lines is a subject ofour current R&D activity.

3. Hle and Fs for shedding calculations are calculated using Chen for gas and liquid frictionfactors, Fi=Fg for interfacial friction factor, and (V,-V,) for inter-facial velocity. This results in slightdifferences between MULTIFLO F, and that used in the shedding spreadsheet.

4. Shedding model assumes that there is the equilibrium stratified flow in the downhill section.This is conservative, unless the slugs are very close together (in which case they wouldn’t be1 :lOOO length). This is because the liquid flowrate required to sustain an equilibrium stratifiedfilm in the downward section cannot be supplied as a steady from an upward sloping line. Theupward line will produce the liquid in slugs, which will decay to give a film. But it is very unlikelythat the 1 :lOOO slug will arrive at the start of the downhill section immediately after other slugs.

i.e. slugs will probably shed more than we predict

5. In determining pigged slug volume, MULTIFLO uses Beggs and Brill for upward sections, andMechan 92 for horizontal and downward. The consistency of this has yet to be verified.

6. Despite 5, we still recommend that the design slug volumes from the logic tree be comparedwith the MULTIFLO pigged slug volume.

7. Whenever slug volumes are quoted they require a slug length and a slug holdup. These valuesare the MULTIFLO predictions. Slug lengths should be used in the assessment of each linesection until the final delivery into a vessel, when length should be converted to volume.

8. All slug volumes at line end are divided by 1.2 to take into account shedding from the back ofthe slug as it enters the vessel.


Section 4. Gas/Condensates

Design of Multiphase Gas and CondensatePipeline Systems

4.1 Introduction

4.2 Basic Design Approach

4.3 Mechanistic Model

4.4 Sphered Liquid Volume

4.5 Fluid Composition

4.6 Flowline Topography


Section 4. Gas/Condensate BP Multiphase Design Manual


LEFT BLANK




4.1 Introduction

This section of the Multiphase Design Manual outlines the basic approach required for thedesign of gas/condensate flowline systems, such as that used for export of gas from the WestSole field to the Easington terminal. The details of the methods available for prediction ofpipeline pressure drop, multiphase flow regime and also total and sphered liquid volumes arealso included.


Section 4. GaVCondensate BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 4. Gas/Condensates

4.2 Basic Design Approach

The criteria on which a gas-condensate system is based are usually a minimum receipt pres-sure at the downstream terminal, the available wellhead (or platform) pressure and the produc-tion profile.

A detailed fluid composition and flowline topography is also required to ensure that reliableresults can be obtained using the design methods detailed in this section.

The following is a step by step approach to the typical calculations and considerations requiredin the design of a gas-condensate system.

4.2.1 Material Selection

The maximum flowrate for any line size in a gas-condensate system is often dictated by erosionrather than pressure constraints. As the erosional velocity limit is dependent on the pipe mate-rial, material selection should be conducted at an early stage. It is based on the fraction of acidgases present as well as the operating pressures and temperatures of the flowline. Typicallygas with 0.2-2.0 per cent CO, at moderate pressures (300-l 500 psia) would be transportedvia carbon steel flowlines with inhibition and a corrosion allowance. Significantly higher CO,content or pressures may require the use of corrosion resistant alloys.

The Materials Engineering Group in the company’s Shared Resource at Sunbury is responsiblefor material selection.

4.2.2 Pipeline Sizing

Once the flowline material has been selected the minimum pipeline diameter can be calculatedfor the given production rates and pressure constraints. This work should be conducted usingthe Mechanistic Model within the BPX MULTIFLO program. This model is available fromVersion 11 of MULTIFLO and is described in detail in Section 4.3.

If the turndown in flowrates over the life of the line is large it may be necessary to constrain theproduction profile in order to reduce the diameter of the flowline. This is because the liquidvolume that can accumulate in the system increases rapidly at low throughputs, leading to thenecessity for frequent sphering or the requirement for large slug catchers.

4.2.3 Erosion Vehcity Limits

Once a flowline diameter and material have been selected the maximum mixture velocity in theflowline must be calculated in order to check that its value does not exceed the erosionalvelocity limit. The mixture velocity calculation is conducted by the BPX MULTIFLO program.The method used to determine the erosional velocity limit is detailed in Section 3.1.4. It is,based on a relationship given in APl14E for droplet erosion. If sand is present, particulateerosion may be of greater significance than droplet erosion. Work, sponsored by BP, hasbeen undertaken at Harwell Laboratories to determine the velocity limits for particulate erosion.The conclusions of this work, and a general discussion on erosion issues, is presented in theBP Erosion Guidlines documents, see Section 3, ref (10)

If the erosional velocity limit is exceeded either the line diameter must be increased or the



production profile constrained to reduce the maximum mixture velocity. Alternatively a highergrade of pipeline material could be used - duplex stainless steel has a higher resistance toerosion than mild steel (Section 3.1.4).

As well as the erosional velocity limit, APl14E also recommends that a line velocity of 60 ft/secshould not be exceeded to ensure that the level of noise emitted by the flow is not excessive.

4.2.4 Slug Catcher Sizing

Once the flowline material and diameter have been chosen, the required capacity of the slug-catcher at the reception facility must be determined. This is based primarily on the volume ofliquid that is produced during sphering of the flowline. This sphered liquid volume is dependenton whether or not the line has reached equilibrium - ie. whether, for the given set of operatingconditions, the total liquid in the line has reached its maximum value.

For non-equilibrium conditions the total liquid volume produced during sphering is assumed tobe equal to the liquid that has entered the line since the launch of the previous sphere. This isdependent on the time interval since the last sphere waslaunched, the gas flowrate and the liquid loading:

Q+tvprod = ~ (non-equilibrium conditions)

24

where:

vprod = produced volume (bbls)Q = gas flowrate (mmscfd)+ = liquid loading (bbls/mmscfd)t = time interval since previous pig launch (hrs)

The liquid loading is the ratio of liquid to gas and varies with pressure, temperature and compo-sition. For this calculation it is normally calculated for the outlet conditions.

For equilibrium conditions the total liquid hold-up in the flowline is calculated using theMechanistic Model in MULTIFLO. The volume of the liquid slug produced during sphering isless than the total equilibrium value as liquid flows out of the line during the passage of thesphere. The calculation procedure for determining the sphered slug volume is given in Section4.4.

The approximate time period required to reach equilibrium conditions is given by:

teqm = ~ (non-equilibrium conditions)Q+

where:

tearn

= time interval to reach equilibrium (hrs)



-

V,, = equilibrium liquid volume (bbls)Q = gas flowrate (mmscfd)4 = liquid loading (bbls/mmscfd)

The sphered liquid volumes for a given flowline system under both the equilibrium and non-equilibrium conditions can be plotted against gas flowrate on a design diagram:

600600 -

- Total Eqm

- - Sph Eqm- - - - - - 24hr

- - - - - 4 8 h r

- - - - 7 2 h r

L-L -

Non-eqm volumesNon-eqm volumes

I200

Gas Flowrate (mmscfd)

3bo

Figure 1.

This figure shows that for equilibrium conditions the sphered liquid volume decreases with gasflowrate, whereas for non-equilibrium conditions, and a set sphering interval, the liquid volumeincreases with flowrate. The maximum sphered volume can therefore be determined at theintersection of the sphered liquid volume lines for the equilibrium and non-equilibrium condi-tions. This is often used as a design basis for determining the required working volume of theslug catcher. The location of this intersection moves to a higher flowrate, and therefore asmaller value of the maximum sphered volume, as the sphering frequency is increased.Typically the intersection between the equilibrium and non-equilibrium lines for sphering intervalof 24 hours will occur at approximately 50 per cent of the design throughput. If the intersectionoccurs at a higher flowrate the line is probably oversized. However, this may be unavoidablegiven the imposed constraints of production profile and allowable pressure drop. At a flowrategreater the value at the intersection, sphering is not necessary to limit the line’s liquid inventory.At flowrates less than that at intersection the maximum time interval between spheres can becalculated from:

24 vsct sph

= (non-equilibrium conditions)Q 4



where:

t = maximum time interval between spheres (hrs)GE = slug catcher working volume (bbls)Q = gas flowrate (mmscfd)Q = liquid loading (bbls/mmscfd)

The design diagram shows that, although the sphered liquid volume tends towards zero as thegas flowrate increases, the sphering operation is still predicted to produce a liquid slug even athigh flowrates. However, field data at high flowrates has indicated that above a certain criticalgas velocity (V,) no slug is produced on sphering. An empirical correlation based on field datagives:

where: -

V, = Critical gas velocity (m/s)D = pipe diameter (m)ps = gas density (kg/m3)p, = liquid density (kg/m3)(T = surface tension (N/m)

The discrepancy between the present model and field measurements at these high flowratesmay be due to the model under-predicting the liquid entrainment in the gas flow (see Section4.3), liquid leakage passed the sphere (see Section 4.4) or an underprediction of the liquid filmvelocity.

4.2.5 Corrosion Inhibition

Corrosion inhibition, using chemicals such as filming amines, is necessary in gas-condensatesystems when the required corrosion allowance for an uninhibited system is excessive. If this isthe case a corrosion allowance for the inhibited system can be calculated from the allowancefor the uninhibited system and the inhibitor efficiency (typically 85 per cent). If the requiredcorrosion allowance is still excessive a higher grade material must be used for the pipeline.Calculation of corrosion allowance and choice of inhibitor is conducted by the MaterialsEngineering Group.

Corrosion inhibitors will only perform at their nominal efficiency if they remain in contact withthe pipe wall. However, under adverse flow conditions inhibitor stripping can occur. For singlephase flow inhibitor effectiveness is only guaranteed in straight, horizontal flowlines below a crit-ical shear stress value. Materials and Inspection Engineering Group in their Corrosion GuidlelineDocument [See Section 3, ref(1 I)] recommend a maximum wall shear rate to avoid inhibitorstripping of 100 N/m2.

The critical shear stress limit is also applied to multiphase systems, where it can be comparedwith a nominal shear stress value determined from the calculated frictional pressure gradient:



D dPf

TV=--4 dx

where:

7, = nominal wall shear stress (Pa)D = pipe diameter (m)

dPf- = frictional pressure gradient (Pa/m)dx

This relationship between shear stress and frictional pressure gradient is based on homoge-neous flow. It always provides a conservative estimate for the actual shear stress ina steady stratified flow - the regime in which gas-condensate flowlines primarily operate.However, the wall shear stress could be substantially higher when slug flow is encountered, onthe outside of bends or on the downstream side of weld beads. To estimate the shear stress-es in these locations computational fluid dynamics using the FLUENT code can be employed.The Multiphase Flow Skills Group have extensive experience in its application.

At present BP has an ongoing programme of research into the factors which dictate corrosioninhibitor performance in multiphase flow. This work may indicate that inhibitor effectiveness isdependent on factors additional to the shear stress.

As gas-condensate pipelines operate primarily in stratified flow regime, corrosion inhibitor isonly deposited on the pipe’s top wall if there is a substantial fraction of the liquid entrained asdroplets in the the gas phase. However, corrosion will still occur due to the condensation ofwater. For an uninhibited system the rate of corrosion at the top of the line will be lower thanfor the of the base of the pipe because, after condensing, the water will rapidly fall back to thestratified liquid layer under gravity. De Waard and Milliams (1) have indicated that a conserva-tive estimate of top of line corrosion is 10 per cent of the overall uninhibited corrosion rate. Thishas been confirmed by experimental work conducted at Sunbury.

A top of line to bottom of line corrosion rate ratio of 0.1, for an uninhibited system, impliesthat corrosion considerations will be dictated by the bottom of line rate for inhibited systems,unless the inhibitor efficiency is better than 90 per cent.

4.2.6 Hydrate Formation

At low temperatures and high pressures water and gas can combine to form hydrate crystals.These crystals can grow to block the flowline. To ensure hydrate formation is avoided, theflowline should be designed to operate outside the hydrate formation region throughout itslength. Insulation requirements to ensure that the line temperatures is maintained outside thisregion can be determined using MULITFLO.

If it is not possible to design to avoid the hydrate formation region, hydrate inhibitors such asmethanol or glycol can be employed. Inhibitors will often be required at low throughput, andalso during a controlled shut-down. If no inhibitor injection is possible prior to shut-down,hydrate plugs may form in the flowline. In these circumstances depressurisation may be



required to ensure the line conditions are outside the hydrate region.

The boundary of hydrate formation region, for static equilibrium conditions, can either be deter-mined experimentally at Sunbury or, for a known fluid composition, using the HYDOl module inBP’s GENESIS flowsheet simulator. Under flowing conditions, the hydrate region will be smallerthan that determined for static equilibrium conditions, and therefore both methods are conser-vative.

4.2.7 Solids Transport

Sand is rarely encountered in gas-condensate systems. However other solids, such as prop-pant, may be present following well fracturing operations.

If solids are present the minimum velocity to ensure their continuous transport must be deter-mined. As, in gas-condensate flowlines, the actual liquid velocity will nearly always be less thatgiven in Section 3.1.4 for solids transport (3 Ws), it is normally assumed that the solids aretransported by the gas phase only. However, the transport of solids in gas-condensatesystems is, at present, little understood. Developments in this field will be included in theSection 12 of this Multiphase Design Manual.

Section 4.2.3 discusses the effect of solids transport on erosion.

4.2.8 S-Phase Flows

In water wet gas-condensate systems it can normally be assumed that the condensate andwater phases are well mixed. However, when the water loading is a substantial proportion ofthe total liquid loading, the water and condensate can separate out at low flowrates leading tothe accumulation of stagnant water at low points in the line. The presence of stagnant watercan cause excessive corrosion at these low points. It could also lead to the production of waterslugs during normal line operation, flowrate changes or sphering.

The conditions under which water drop-out occurs at low points in flowlines need to bepredicted to assess the potential for corrosion and correctly design water handling facilities.The required modelling techniques for 3-phase flows will be discussed in Section 14 of theMultiphase Design Manual.

4.2.9 Transient Operation

Sections 4.2.1 to 4.2.8 discuss the design of gas-condensate flowlines based on operationunder steady state conditions. However, the response of the system to start-up, shut-downand flowrate changes must also be considered. For example, liquid slugs are often producedduring flowrate increase or line depressurisation. The volume of these transient slugs must bedetermined to ensure that they can be handled by the downstream separation and processingfacilities. If there is insufficient capacity, operating procedures can be introduced to limit therate of change of flowrate and therefore the size of any produced slugs.

This type of transient analysis can be conducted using transient simulators, such as PLAC, thatare currently under development. A full discussion on the use of transient simulators will beincluded in Section 7 of the BP Multiphase Design Manual.


BP Multiphase Design Manual1


The compressibility of the fluids in gas-condensate systems can also lead to linepack. For agiven outlet pressure the average flowline pressure, and therefore the line’s capacity to storegas, increases with flowrate. Thus, an increase in inlet flowrate leads initially to linepack withno immediately impact on the flowrates at the outlet of the line. Equally, if the inlet flowrate isreduced, line depressurisation ensures that the outlet flowrates can be maintained for a periodof time that depends on the overall line volume and operating pressure. Linepack or linedepressurisation also occurs when the outlet pressure is changed but the inlet flowrate is keptconstant.

If the line capacity is sufficient, linepack can be used, in combination with pressure variationsat the flowline outlet, to allow steady conditions to be maintained at the inlet whilst supplyingthe fluctuating daily demand for gas. It also allows the gas supply to be maintained even if theproduction facility is shut down for short periods.

Line pack and depressurisation can be studied using the TGNET transient code. TheMultiphase Skills Group maintain a license to use TGNET and have extensive experience in itsapplication.

issue 2 - December 1994 I I



LEFT BLANK




4.3 Mechanistic Model

This section describes the Mechanistic Model for the prediction of pressure drop and total liquidhold-up for systems that have reached equilibrium and are operating under steady state condi-tions. This model is included within the BPX MULTIFLO program as “Segregated Flow - GREmechanistic model” (1992). The method is used to initially determine the flow regime in eachsection of the flowline. The calculation procedure for pressure drop and liquid hold-up (theproportion of the flowline volume or cross-sectional area occupied by the liquid phase) isdependent on this predicted flow regime.

4.3.1 Flow Regime Determination

The flow regime at any point in the flowline is determined primarily using the Taitel-Duklermethod (2). This model identifies the following five flow regimes:

(a) Stratified-smooth(b) Stratified-wavy(c) Intermittent (or slug)(d) Annular-mist(e) Dispersed-bubble

The determination of flow regime is conducted in two stages.

Initially stratified-smooth flow is assumed and an equilibrium liquid level is calculated. This isthe stable liquid level that would occur if the flow were stratified with a smooth interface and nodroplet entrainment in the gas-phase. It is determined by equating the pressure gradient in theliquid and gas layers.

Once the equilibrium liquid level has been calculated the stability of the liquid film is considered.At low flowrates no waves form on the film surface and the stratified-smooth flow is stable. Athigher flowrates stable waves may form which do not grow to block the pipe - the stratified-wavy regime. If unstable wave growth does occur, slugs will form and the flow regime is inter-mittent. At high gas and low liquid flowrates there may be insuflicient liquid in the system forslugs to form. In this case unstable wave growth leads to annular-mist flow. Finally for highliquid rates and low gas rates the gas is entrained as bubbles within the liquid flow -thedispersed-bubble regime. The Taitel-Dukler method models these mechanisms in order todetermine the liquid and gas superficial velocities at which each of these transitions occur. Theflowrate at the transition is a function of the fluid properties, the flowline diameter and the flow-line inclination.

For the mechanistic model several modifications have been made to the original work of Taiteland Dukler. The method for determining the position of the intermittent/annular boundary hasbeen modified following work by Barnea (3). Also, when determining the equilibrium liquid level,the interfacial shear stress term is calculated by setting the value of the inter-facial velocity to beequal to the liquid film velocity and not zero. Finally, work by Landman (4) has shown that thereis more than one possible value for the equilibrium liquid level under certain special conditions -ie. low liquid loadings and upwardly inclined flowlines. If this is the case the mechanistic modeluses the smallest of the three values, as this is considered to be the stable solution for realsystems. The predictions of the stratified/intermittent boundary position may also be improvedin the future following work conducted at BP Research (5).



The method of Taitel and Dukler has also been extended to include two additional flow regimes:(f) Liquid(s) Gas

and divide the intermittent regime into two sub-regions:

(cl) Normal slug(~2) Terrain-induced slug

The liquid and gas flow regimes are simply used to identify single phase gas and liquid flows.These are defined to be when the no-slip liquid hold-up is greater than 0.9999 or less than0.0001. The no-slip liquid hold-up is the volume proportion that would be occupied by theliquid if the flow were fully mixed.

The division of the intermittent regime is only relevant for sections of flowline with positive(upward) inclinations. The influence of the gravity term in uphill sections of flowline leads to theformation of slugs at much lower flowrates than would be the case for a horizontal line.Whereas slugs formed at higher flowrates (normal slugs) will be relatively unaffected by changesin flowline inclination, those formed at low flowrates in uphill sections of flowline (terrain-inducedslugs) will not be stable in horizontal or downhill sections. Terrain-induced slugs are also char-acterised by the counter-current nature of the film that is shed from the rear of each slug. Thedifferences between normal and terrain-induced slug flow leads to the necessity to employseparate methods for the calculation of liquid hold-up and pressure drop.

Gas/condensate systems generally operate in the stratified-smooth, stratified-wavy, annular-mist or gas flow regimes. Terrain-induced slugging may also be encountered in uphill sectionsof flowline at low flowrates. The remaining flow regimes (normal slug, dispersed bubble andliquid) are only encountered in black-oil systems.

4.3.2 Pressure Drop/Liquid Hold-up Calculations

(a) Stratified and Annular Flow Regimes

The approach used for calculating the pressure drop and liquid hold-up in the stratified-smooth,stratified-wavy and annular-mist flow regimes is identical. In each case a segregated flow isassumed and the liquid hold-up and pressure drop are determined using a similar calculationprocedure to that employed when determining the equilibrium liquid level for flow regime identi-fication (Section 4.3.1).

However, for stratified-wavy or annular flows the assumptions of a smooth interface and noliquid entrainment in the gas phase are not correct. To take account of the effects of inter-facialwaves and droplet entrainment the pressure drop equations in the gas and liquid layers havebeen modified.

Inter-facial waves are accounted for by using the inter-facial friction factor equation of Andreussiand Hanratty (6). This leads to the prediction of greater losses through an increased inter-facialstress term. The overall effect of using the Andreussi-Hanratty equation is to increase thepredicted pressure drop and decrease the predicted liquid hold-up at high flowrates.

The entrainment of liquid droplets in the gas phase can be accounted for by treating thegas/droplet mixture as a single modified gas phase. The density and viscosity of this



gas/droplet phase are calculated assuming that the mixture is homogeneous:

p* = c, p, + (1 - Cd) P,

P!F = c, PI + (I- C,) P,

where:

C, = droplet concentrationp, = gas density (kg/m3)p, = liquid density (kg/m3)

pw = density of gas/droplet mixture (kg/m3)p, = gas viscosity (kg/m/s)p, = liquid viscosity (kg/m/s)IJ- ge = viscosity of gas/droplet mixture (kg/m/s)

In these equations the droplet concentration is the ratio of the cross-sectional area occupied bythe liquid droplets to that occupied by the gas/droplet mixture. It is related to the entrainmentfraction (the ratio of the liquid mass flux transported as droplets to the total liquid mass flux) asfollows:

where:

E, = entrainment fraction

Vsl = liquid superficial velocity (m/s)

vsg = gas superficial velocity (m/s)

The treatment of the gas and droplets as a single phase is based on the following assumptions:

l The droplets are transported at the velocity of the gas. Thus drag between the gas and thedroplets can be ignored.

l There is no momentum transfer between the gas/droplet phase and the liquid layer due tothe droplet entrainment and deposition processes.

l The droplet entrainment and deposition processes do not need to be modelled separatelybut can be accounted for within the calculation of the entrainment fraction.

Unfortunately there is little entrainment fraction data available for horizontal stratified and annularflow. The method used within the mechanistic model is based on data obtained by Shell ontheir high pressure Bacton flowloop (7). There is evidence that this method may under-predictthe entrainment fraction at high gas flowrates.

The inclusion of entrainment fraction also leads to an increase in pressure drop and a decreasein the liquid hold-up predictions.

Issue 2 - December 1994 I5


(b) Gas, Liquid and Dispersed Bubble Flows

The calculation of the liquid hold-up and pressure drop in the liquid, gas and dispered bubbleflow regimes is conducted by assuming the flow is a homogeneous mixture. The flow is treatedas a single-phase with the mixture density and viscosity:

“,I PI + “sg pg “4 h + “sg &IPm = and km =

“,I + “sg “,I + “sg

where:

P, = mixture density (kg/m?p, = gas density (kg/m3)pl = liquid density (kg/m3)

kT = mixture viscosity (kg/m/s)ps = gas viscosity (kg/m/s)~1 = liquid viscosity (kg/m/s)VSI = liquid superficial velocity (m/s)V

%I= gas superficial velocity (m/s)

(c) Normal Slug Flow

At present no mechanistic model is available for accurately calculating the pressure drop andhold-up in normal slug flow. Thus, the empirical method of Beggs and Brill is used whennormal slug flow is predicted (see Section 3.1.3).

(d) Terrain-Induced Slug Flow

If terrain-induced slug flow is predicted the hold-up and pressure drop are calculated from themaximum stable liquid accumulation determined in Appendix 3B.

This maximum stable liquid accumulation is the maximum volume of liquid that can be retainedin an uphill section of flowline at a given gas velocity. It is calculated by assuming that slugsform at the start of the uphill section of flowline, pass through the line shedding liquid from theirrear and then collapse just before they reach the end. The shed film forms a counter-currentliquid film which flows back to the start of the uphill section of flowline where it builds up to formthe next slug.

Liquid flowing into the uphill section of flowline leads to the continuation of short slugs into thenext section. However, the liquid retained in the uphill section remains equal to the maximumstable. liquid accumulation.

For terrain-induced slugging the liquid hold-up in an uphill section of flowline is simply equal tothe ratio of the maximum stable liquid accumulation to the total volume of the pipeline section.

From Appendix 3B this is:


BP Multiphase Design Manual Section 4. 4. Gas/Condensates

0.16~~0.16~~ + 0.59 + 0.59H, H, == HISHIS

l.l6v,l.l6v, + 0.59 + 0.59

where:

H, = liquid hold-upH,, = slug hold-up

“n- = mixture velocity (m/s)

The slug hold-up is calculated using a steady state correlation - at present the method ofGregory is employed (Appendix 3A).

The calculation of pressure drop for terrain-induced slugs assumes that the flow is homoge-neous with the mixture density and viscosity based on the calculated hold-up:

Ptll = h, P, + (1 - HJ P,

kl = H, = H, p.,p., + + (1(1 -- H,)H,) CL~CL~

This approach ensures that a good approximation to the hydrostatic component of the pressure gradient is obtained. This component is normally much greater than the frictional compo-nent when terrain-induced slugging is encountered.




LEFT BLANK



;I


4.4 Sphered Liquid Volume

As discussed in Section 4.2.4 the sphered liquid volume, for a flowline that has not yet achievedequilibrium, can be estimated as the total volume of liquid that has entered the flowline followingthe launch of the previous sphere.

For the equilibrium case the sphered liquid volume is equal to the total liquid volume in the line,calculated using the Mechanistic Model, less that swept out of the pipeline as a liquid filmduring the sphere transit time. To determine the sphered liquid volume the growth of the slug infront of the sphere must be calculated, using the mass balance equation, throughout thesphere’s passage through the line.

The process involved with the passage of a sphere through the flowline is shown below:

Figure 2. Growth of liquid slug during sphering growth

where:

H, = liquid film hold-upH,s = sphered slug hold-upL, = sphere to pipe cross-sectional area ratiova = velocity of liquid at sphere (m/s)v, = liquid film velocity (m/s)V = mixture velocity (m/s)VI = sphered slug fluid velocity (m/s)Vsph = sphere velocity (m/s)v, = sphered slug front velocity (m/s)

The mathematical details of the approach required to calculate the sphered slug volume will notbe discussed in this design method. However, the general approach is to determine the rate ofgrowth of the sphered slug as it passes through the flowline collecting liquid from the precedingfilm. Liquid leakage passed a sphere that does not fully block the pipe can also be included.

Thus, a combination of three equations for:

(a) The mass balance at the front of the growing slug

(b) The mass balance at the sphere (if liquid leakage passed the slug ismodelled)

(c) The growth of the slug due to the differential velocity between the slugfront and sphere

Issue 2 - December 1994 I9


lead to the required method for integrating the slug length as the sphere passes through theflowline.

For the special case of a horizontal flowline, zero pressure drop, no liquid leakage passed theslug and a sphered slug hold-up equal to unity, this equation reduces to:

where:

Vtot = total liquid holp-up in the line prior to spheringVout = liquid volume leaving the line during passage of sphere

This approximate form for calculating the sphered slug volume has been employed in the BPXMULTIFLO program.

-

2 0 issue 2 - December 1994


4.5 Fluid Composition

The calculation of equilibrium and sphered liquid volumes and, to a lesser extent, that of thepressure and temperature drop in gas-condensate systems is very sensitive to fluid composi-tion. For instance, the liquid loading of the line may double due to small changes in the propor-tion of the heavier components, leading to an increase in the predicted sphered slug volume ofmore than a factor of two.

Given the importance of using the correct fluid composition, black oil correlations are not suit-able for gas-condensate design. Instead the FLODAT module of the GENESIS flowsheet simulator should be used to produce a compositional datapack which is then accessed by MULTI-FLO.

To ensure that the correct pipeline composition is used for the production of the compositionaldatapack any saturated water and offshore processing must be taken into account.

4.5.1 Water Content

Normally the compositional data will be available for the dry wellstream only. However, the gasis normally saturated with water at reservoir conditions. To determine the actual water contentof the fluid, GENESIS can be used to add water to the dry wellstream composition. By usingthe ASAP program in GENESIS, this wet composition can then be flashed at reservoir condi-tions, to determine the actual water content of the gas.

Although most gas wells have no free water at reservoir conditions, later in field life water break-through can occur and free water must be taken into account.

If the gas is dehydrated offshore using a glycol contactor, negligible water remains and the drywellstream composition can be used.

However, if free water knock-out is used, the water fraction in the flowline is the saturated watercontent at the pressure and temperature of the water knock-out equipment.

4.5.2 Addition of Inhibitors

The liquid loading of a gas-condensate flowline can be substantially increased by the addition ofinhibitors at the upstream end of the flowline. To ensure the correct liquid loading is used forgas-condensate calculations these inhibitors must be added to the composition. Unfortunately,GENESIS cannot correctly partition corrosion inhibitors between the oil and water phases. Itmust therefore be assumed that the inhibitors are contained entirely within the water phase.This can be carried out by substituting water for the inhibitor in the same volume ratio to thegas, and adding this to the composition determined in 4.51 by using ASAP.

4.5.3 Combined Compositions

Obviously if the production from more than one field is to be transported by a gas-condensateflowline, the combined composition must be used. Thus, compositional datapacks must beprepared for each ratio of flowrates from the two fields.




LEFT BLANK




4.6 Flowine Topography

The predictions of the Mechanistic Model, and also a second simpler mechanistic type modelcalled Segreg, can be sensitive to the flowline topography at low flowrates. This sensitivity is areal effect and cannot be ignored. Thus, although empirical methods, such as those Oliemansfor pressure drop and Eaton for liquid hold-up, are not sensitive to the line topography, this isdue to their inability to correctly model the changes in flow regime and liquid level between hori-zontal and uphill sections of pipeline.

To ensure that the predictions of liquid hold-up and pressure drop are accurate it is thereforeimportant to ensure that as detailed a topography as possible is used. It is generally the casethat the liquid hold-up will increase asymptotically to the final value as the number of linesegments is increased. Thus, calculations based on a simple approximation to the topographywill usually lead to an under-estimate of the liquid hold-up.

The errors incurred by using an over-simplified approximation to the real topography will not besignificant for flowlines in which inclinations are small (of the order of kO.1 ‘, or if gas velocitiesare high (greater than 30 ft/s). However, if large inclinations or low gas velocities are encoun-tered the errors can be large. An example of this is the North East Frigg flowline. The depth ofthis flowline varies by up to 90 feet over its 11 mile length, with inclinations of between +3.7”and -3.7”. Three topographies were used to model the flowline section, consisting of 1, 9 and18 pipeline segments respectively. For the lowest flowrate, giving a gas superficial velocity ofapproximately 2.5 f-t/set, the calculated pressure drops for these three topographies were 35,57 and 74 psi respectively and the liquid hold-up values were 101, 1284 and 2496 bbls. Themeasured values, which are themselves only approximate, were 72 psi and 2410 bbls - inclose agreement with the most detailed topography used.

It is possible that the predictions would increase further if an even more accurate topographywere available. However, as the 18 segment topography covered all the major inclinationchanges, any further increases in the predictions are likely to be small.

For reference the Olieman pressure drop correlation combined with the Eaton hold-up correla-tion predict a line pressure drop of 40 psi and liquid hold-up of 500 bbls with the detailedtopography.

The North East Frigg example is an extreme case of low flowrate and large upward inclinations.However, it leads to the recommendation that all topography changes which result in a pipelinesection that slopes upwards at more than +O.l’ should be included in the topography toensure reliable results.




LEFT BLANK



4.7 Slug Flow in Normal Operation

Section 4.4 detailed the method employed for calculating the volume of a liquid slug producedon sphering. Slugs may also be produced through the outlet of the line without sphering.These slugs can be caused by changes to the operating conditions of the flowline. If this is thecase, the transient simulators briefly mentioned in Section 4.2.9, and discussed in detail inSection 7, can be used to predict the volume of slugs produced in this way.

Slugs can also be produced under steady state conditions if the flowline is predicted to beoperating in the slug flow regime. For systems with a high liquid loading (at least 100bbls/mmscf) normal slug flow is possible. In this case slugs can form in horizontal sections,and are likely to persist throughout the length of the flowline. The BPX slug flow model inMULTIFLO should be used to predict the length of these normal slugs. This method is detailedin Appendix 3A.

When terrain-induced slugging is predicted, slugs will only form in upwardly inclined section offlowline. These slugs are unlikely to persist throughout the length of the line. Instead they willsteadily decay and then collapse in horizontal or downwardly inclined sections. However, theproduction of terrain-induced slugs at the outlet of the line may occur if the flowline is inclinedupwards to the reception facilities. This is often the situation for lines between offshore fieldsand coastal reception terminals. For these circumstances the method based on slug genera-tion in dips outlined in Appendix 3B of this manual should be used.


ISection 4. Gas/Condensate BP Multiphase Design Manual


LEFT BLANK



References

1. De Waard, C, Millams, D E, Predictive Model for CO, Corrosion Engineering in Wet Natural GasPipelines, paper presented at The NACE Annual Conference and Corrosion Show, Cincinatti,March 1991.

2.

3.

4.

5.

6.

7.

Taitel, Y, Dukler, A E, A Model for Predicting Flow Regime Transitions in Horizontal and NearHorizontal Gas-Liquid Flow, AlChE Journal, 22,47-55, January 1976.

Barnea, D, Transition from Annular Flow and from Dispersed Bubble Flow - Unified Models forthe Whole Range of Pipe Inclinations, Int J Multiphase Flow, 12, 733-744, 1986.

Landman, M J, Hysteresis of Holdup and Pressure Drop in Simulations of Two-Phase StratifiedInclined Pipe Flow, Multiphase Production, published by Elsevier Science Publishers, pp 31-50,1991.

Wood, D G, Slug Characterisation Studies - Experimental Studies on a Horizontal 6-InchAir/Water Flowloop, RCS Branch Report No 124335 dated 06.04.92.

Andreussi, P, Persen, L N, Stratified Gas-Liquid Flow in Downwardly-Inclined Pipes, Int JMultiphase Flow 13(4), pp 565-575, 1987.

Pots, B F M, Maui-B Satellite Study - Corrosion Aspects of Carbon Steel Pipelines - Analysis ofThree-Phase Calculations, Shell Note EV426/88, September 1988.




LEFT BLANK

2 8 issue 2 - December 1994

BP Multiphase Design Manual Section 5. Oil Line Water Slugs

Section 5. Oil Line Water Slugs

Calculation of Pigged Water Slug Volumesfrom Oil Lines

5.1

5.2

5.3

5.4

5.5

Introduction

Nomenclature

Equilibrium Hold-up

Slug Volume

Conservatism

Appendices


Section 5. Oil Line Water Slugs BP Multiphase Design Manual


LEFT BLANK



5.0 Foreword

This section of the Multiphase Design Manual describes a method for calculating the volume ofa water slug produced when pigging an oil line, which contains small quantities of water. Themethod assumes that the water forms a segregated layer at the bottom of the pipe. Thisassumption is valid for most oil transport lines but under some high flowrate conditions at leastsome of the water will be carried along as entrained droplets. Under these circumstances themethod given in Section 5.4 below will over predict the pigged water slug volume.

A second calculation procedure has been developed which checks if water will form a segre-grated layer. It determines the water concentration profile in oil/water pipelines and estimatesthe water layer depth.

This method has been incorporated into a computer program called WATPROF, which is avail-able from the BPX Multiphase Flow Group. It is discussed in detail in Appendix 5A.

The following input data is required for WATPROF.

DEU’C

J-42POPWCT

Pipeline diameter (m)Pipe roughness (in)Mean fluid velocity (m/s)Water cut (%)Oil viscosity (Ns/m2)Oil density (kgIm3)Water density (kg/m3)Surface tension (N/m)


Section 5. OA Line Water Slugs BP Multiphase Design Manual


LEFT BLANK



5.1 Introduction

This Multiphase Design Manual describes the phenomena of water hold-up in water/oil multi-phase lines. This note details the steps required for the calculation of the slug volume of awater slug resulting from the pigging of the line.

Free water will tend to be collected at all points in the line at which the velocity of the waterphase is retarded relative to that of the oil phase. This behaviour, termed ‘hold-up’, is reflectedas an increase in the in-situ volume fraction of water (H,) above its flowing water fraction (C,).Hold-up principally occurs under conditions of stratified flow where the water forms a distinctlayer at the bottom of the pipe. Where the concentration of water is relatively low, the separat-ed water resides primarily within the pipe boundary layer. As a result, its average velocity is lessthat that of the overlying oil. This velocity difference is further increased by the greater inertialstresses acting on the water phase. The presence of uphill sections within a pipeline will furtherenhance the velocity differences, whilst downhill sections will reduce them. since the pig travelsat the average velocity of the line the mechanism of water collection may be likened to one inwhich the pig ‘overtakes’ the slower moving layer.


Section 5. Oil Line Water Slugs BP Mu&phase Design Manual

.-


LEFT BLANK



5.2 Nomendature

The following nomenclature is used for the explanation of the calculation to estimate the waterslug volume:

QcbvHbvtALv

L

Oil flowrate (m3/s)Inlet water contentEquilibrium hold-upTime since last pig (s)Area of line (m2)Length of line (m)Slug volume (m3)Time for the line to reach equilibrium (s)




LEFT BLANK



5.3 Equilibrium Hold-up

The water hold-up fraction is influenced by the density and viscosity of both liquids, flowingwater fraction, the pipe diameter, mixture flowrate, and the pipe inclination. Of these, the pipeinclination and the flowing water fraction are the most important.

Water hold-up increases with the flowing water fraction and the pipe inclination. As a generalrule, when the flowing water fraction increases over 10% the flow regime is no longer stratifiedand the hold-up will tend towards the flowing water fraction. Water slugs will be created byaccumulation and sweep-out, the size of which will mainly depend on the geometry of the line.

The calculation of the hold-up is complicated and is not described in this document but software is currently under development to perform this calculation.


Section 5. Oil Line Water SLugs BP MuEtiDhase Design Manual

‘-


LEFT BLANK



5.4 Slug Volume

5.4.1 Whole Line at Equilibrium

The main assumption that all these calculations are based on is that no water leaves the lineuntil the whole line has reached equilibrium. Therefore once equilibrium has been reached, thetotal volume of water in the system is;

AxLxHW Equation 1

At the moment the line reaches equilibrium (the moment water starts to flow from the end of theline) the total water entered since the last pig will be:

QxC,xT, Equation 2

Therefore:

AxLxHW = QxC,xT, Equation 3

AxLxHwT, = Equation 4

QxCw

The volume of water produced as a slug will be the water content less the volume of waterwhich flows out naturally:

V = AxLxHw-QxCwxt” Equation 5

t” = the time it takes for the pig to pass through the line.

volume of line AxLt” = =-------

volumetric flowrate Q

Therefore:

AxLv= AxLxH, -QxC,x ~

Q

V = AxLx(HW-C,)

Equation 6

Equation 7

Equation 8

Issue 2 - December 1994 II

Section 5. Oil Line Water SZugs BP Mu&phase Design Manual

5.4.2 Line Yet to Reach Equilibrium

If the line has not yet reached equilibrium but if a volume of water with a volume greater than Vhas entered the line the slug volume will be the same as V. This is because no water can flowout of the line until the whole line has reached equilibrium. Therefore even as the pig is travel-ling along the line building up a water slug in front, new parts of the line will still be reachingequilibrium nearer to the exit of the pipe.

If the volume of water that has entered the line is exactly V when the pig is launched, then nowater will leave the pipe until the slug (with volume V) arrives.

The time taken for the ‘equilibrium slug volume’ (V) to flow into the line is given as;

AxLx(H,-C,)T, = Equation 9

QxC,

Ift<T,then;

V = QxC,xt

Ift>T,then;

V = AxLx(H,-C,)

5.4.3 Consecutive Pipelines

Equation 10

Equation 11

Consecutive pipelines means a system where one pipeline (the upstream pipeline) flows direct-ly into a second pipeline (the downstream pipeline). It is assumed that the only productionentering the downstream line is from the upstream line. It is assumed that a pig is not launchedinto the downstream line until a pig has arrived in the pig reciever at the end of the upstreamline.

(a) Both Lines Pigged

If the pigging philosophy is such that when one pig arrives at the end of the first line another islaunched into the second, then the calculation of slug volumes is simple. Since the calculationis based on volume, the area and length of the first and second line (A,L, and AJJ can beaveraged to produce an average area and length (Aave and L,J. Therefore:

Awe x Law x (Hw - CJT, =

QxC,Equation 12

If t < T, then:

v= QxC,xt Equation 13



If t > T then:

v = A,, x L,, x W, - C,J Equation 14

(b) First Line Pigged Only

If there are two consecutive lines and only the first is pigged, then the calculation of the pig sizeis based on only the first line in isolation of the second. Any slug which results from the piggingof the first line is assumed to pass through the second line unaffected. Any accumulation ofwater at the front of the slug of water as it travels along the second line is matched by thedeposition of the water behind the slug. The equations 10 and 11 apply with L and A equal tothe length and area of the first line.

(c) Second Line Pigged Only

If the second line is pigged without the first being pigged then the assumption that no waterleaves the line before the whole line reaches equilibrium is no longer conservative. It should beassumed that no water is held up in the first line and again equations 10 and 11 apply with Land A equal to the length and area of the second line. This is conservative but without morefield data it is the only practical method of determining the water slug volume.




LEFT BLANK



5.5 Conservatism

The assumption that no water leaves the line until equilibrium has been reached is conservative.In practice, water will flow from the line before equilibrium is reached, but it is assumed that theline reaches equilibrium as soon as sufficient water has entered the system. This results in theconservative (i.e. high) estimation of the water hold-up in the line and therefore conservative (i.e.again high) prediction of the water slug volumes when the line is pigged.

This is illustrated below in Figure 1, where pigged slug volume is plotted against time since lastpig launch. The hard straight line illustrates the conservative assumption made here, ie liquidaccumulates in the line, and does not exit the line, until the equilibrium slug volume (AL (H&Jhas built up in the line in time T,. In practice it takes some time greater than T, to accumulatethis volume of liquid, as shown by the dashed curve.

If there is a significant period of time between each pig and the line has been at equilibrium for asignificant period then the estimation of the slug volumes may not be so conservative. This isbecause the estimation of the equilibrium hold-up is not excessively conservative.

AL(’ ’_--- - - -/** - - - - - - -

/*-

T, (Time)

Te = Time for pipe (length L) to reach equilibriumAL = Volume of pipeH,,,, = Equilibrium hold-upCw = Inlet water fraction

Figure 1. Pigged slug volume




LEFT BLANK

I6 issue 2 - December 1994


5A Appendix: The Water Concentration ProfIle

5A.l Nomenc&ure Used

C(Y)d95

fKU*

U’

V’

w

Y’5

E(k)

Water concentration at depth y from centreline. (dim.)Droplet diameter corresponding to 95% volume of droplets smaller than d,,. (m)Darcy-Weisbach friction factor.Parameter: (- w/&r.). (dim.)Friction velocity: (= u’(f/8)“.5). (m/s)Mean fluid velocity. (m/s)Water:Oil ratio.Water settling velocity. (m/s)Reduced vertical coordinate, y/R. (dim.)Dimensionless diffusivity in pipe flow.Mathematical function dependent on k.

5A.2 Introduction to Karabelas

In steady horizontal pipe flow, the density difference between the dispersed solid or liquidphase and the continuous fluid phase can cause a non-uniform distribution of the dispersion inthe pipe cross-section. This is the case for lines carrying oil / water mixtures. Karabelas (1977and ‘78), looked at a vertical distribution of dilute suspensions in turbulent pipe flow and atdroplet size spectra, generated in turbulent pipe flow of dilute liquid / liquid dispersions.

The WATPROF program uses some of the results derived in these papers. This appendix brieflyexamines the calculation method and assumptions made throughout.

54.3 The Calculation and Assumptions

Karabelas (1977), obtained his experimental data for (Equation 1) using dilute suspensions ofspherical particles, typically of about 0.5mm in diameter. This is applicable to oil / water disper-sions, because usually there is a low concentration of water and consequently a droplet size ofonly a few millimetres at the maximum, which suggests the droplets will be approximatelyspherical.

The solution for the concentration distribution can be reduced to:

E(K)c(y) = 1 + - x exp(Kxy’)

V’Equation 1

if the droplet size is nearly uniform.

In this expression K = W&J* by definition. This parameter, and especially u* essentially deter-mine the magnitude of the vertical concentration gradients, with a high value of K indicating thatwater will tend to settle out.

The term E(K) depends only on K, and its determination is given in Karabelas(l977). V’ is theratio of water to oil in the fluid.



This approach does not take into account droplet coalescence. Hinze(l955), states that whenthere is a low concentration of the dispersed phase the chances of coalescing are small. Whilsthe does not define ‘low concentration’, droplet coalescence should not be significant in mostlines where the water concentration is less than 1 &15%.

The program considers the water layer to be continuous when c(y) 5 0.52. This value of 0.52volume fraction is the closest packing density of spheres in a cubic lattice. This will obviouslyresult in an over estimate in the depth of the water layer.

5A.4 Other Factors to Consider

(a) The droplet size, d,,, is calculated from a result in Karabelas (1977), and is needed tocalculate the water settling velocity. Although this is based on an isotropic andhomogeneous flow field, the results yielded are accurate enough for this application. Indeed,a sensitivity check showed that a 1% change in d,, would result in a change of 0.7%change in K. The effort required in obtaining a better value can not be justified in this case.

(b) In order to determine the shear velocity, u*,the friction factor has to be calculated. The Chenequation is used instead of the iterative Colebrook-White implicit equation to do this. It givessatisfactory results over the full range of Reynolds numbers and relative pipe roughnesses.

(c) The value of the dimensionless lateral particle diffusivity, 5 ,is set to 0.25. This is taken fromKarabelas (1977) and it is the average value from his experimental work. This is asatisfactory estimate for this calculation.

5A.5 Summary

This model is suitable for oil/water mixtures flowing in long horizontal, or near horizontal,pipelines in a steady state. It is most accurate for small water cuts but still satisfactory up to15-20%.

No account is taken of the increase in oil velocity that is expected if the water layer is deep, soagain the estimate is conservative. Despite this WATPROF is a useful program giving a goodindication of water depth in theory.

5A.6 References.

Karabelas,A. J.“Vertical Distribution of Dilute Suspensions in Turbulent Pipe Flow.”AIChemE Journal. (Vol.23, No.4) July 1977.Karabelas,A.J.“Droplet Size Spectra Generated in Turbulent Pipeflow of Dilute Liquid/liquid Dispersions.”AlChemE Journal. (Vol.24, No.2) March 1978.

Hinze,J.O.“Fundamentals of the Hydrodynamic Mechanism of Splitting in Dispersion Processes.”AlChemE Journal. ( Vol.1, No.5 ) September 1955.


BP Multiphase Design Manual Section 6. Wellhead Temperatures

Section 6. Wellhead Temperatures

Prediction of Wellhead Temperatures underSteady State and Transient Conditions

6.1 Introduction

6.2 Steady-State (MULTIFLO)

6.3 Calculation of Wellhead Temperature

6.4 Unsteady-State (WELLTEMP)

6.5 Appendix

Section 6 Wellbead Tempera tu ws BP Multiphase Design Manual


LEFT BLANK



Section 6 Wellbead Temperatures

6.1 Introduction

This section of the Multiphase Design Manual describes why the accurate calculation of theWellhead Flowing Temperature is important, it also outlines how the WHFT is calculated byBP’s steady-state flowline and well simulation program, MULTIFLO. Section 6 is split into twoparts, the first describes the factors that influence the WHFT and the theory involved, thesecond section (Appendix 6A) describes the more practical aspects involved in determining thewell head flowing temperature using MULTIFLO.

It is important that the temperature at the wellhead is calculated accurately since it can greatlyeffect the design of the flowline and/or the separator downstream of the wellhead. Forexample, the following are influenced by the wellhead temperature:

l The relative vapour/liquid split in the separator or flowline (which may affect the sizing of theflowline or the separator).

l The possibility of hydrate formation.

l Rate of corrosion in the flowline.

l The temperature profile at the inlet to the flowline which effects the type of insulation andburial requirements. If the temperature is high then burial may be required to preventupheaval buckling.

This design guide discusses the factors which affect the wellhead temperature, such as watercut, flowrate, and completion details. In addition, it summarises the calculation used by the twoprograms which are most commonly used within BP to calculate Well Head Flowing tempera-ture: MULTIFLO and WELLTEMP:

l MULTIFLO is BP’s steady-state two-phase simulation program. It does not rigorously modelthe heat transfer within the tubing and the casings. Instead, the casings and annuli areapproximated to an equivalent thickness of cement (or another insulation material). Thismethod provides accurate results considering that information on the heat transfer throughthe completion details and the thermal properties of the surrounding rock may not be readilyavailable and/or of questionable accuracy. MULTIFLO is BP’s preferred method of calcu-lating steady-state wellhead flowing temperatures.

l WELLTEMP is a unsteady-state well simulation program. It rigorously models the tempera-ture loss through the tubing, casing and the annuli with time. It can calculate the tempera-ture profile in the well during start-up and as a result of flowrate changes. WELLTEMPshould not be used to calculate steady-state wellhead flowing temperatures butshould only be used to give an indication of the warm-up rates or the thermalresponse to changes in the operating conditions such as flowrate.


Section 6 Wellbead Temperatures BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 6 Wellhead Temperatures

6.2 Steady-State (MULTIFLO)

6.2.1 Introduction

This section describes the method used by MULTIFLO to calculate the WHFT. The descriptionof the calculation will allow the reader to determine which factors most influence the WHFT.

6.2.2 Calculated Well Heat Transfer

The heat lost from a flowing well into the surrounding rock matrix is by time dependent heattransfer. As in the case of a flowline, the overall heat transfer coefficient, which determinesheat loss from the produced fluids, is obtained from a series combination of three components:

l An Internal Film Coefficient

l A Composite Coefficient for the Tubing/Casing Comprising:- the resistance of the tubing and casing walls- the resistance of the annulus medium- the resistance of the casing cement grouting

l A time dependent coefficient for the surrounding rock matrix.

This summary does not include the contribution to the overall heat transfer coefficient of theheat transfer coefficient for radiation as it’s influence is usually small compared with theother coefficients.

6.2.3 Heat Transfer Through Completion

The calculation of the heat transfer rate through the layers becomes quite complicatedwhen a large number of casings are employed. The numerous casings, annuli and cementthicknesses each contribute to the overall resistance to heat transfer. To simplify thecalculation, it is possible to approximate the numerous layers to a single layer of cementhaving the same resistance to heat transfer as the sum of the casings and annuli.

The heat transfer rate, Q, is described below. Where ‘j’ indicates the properties of the jthlayer.

27ATQ= Equation 1

f, [In (D,+, 1 Di) 1 ki]

For a single layer of cement with a outer diameter of D,, the equation becomes:

2xAT 2xkcATQ= = Equation 2

In (DC / D,i) / k, In (DC/ D,i)


Section 6 Wellhead Temperatures BP Multiphase Design Manual

Since DC is the only unknown and the numerator is the same for both equations theequations can be rewritten as below:

D, = Dci exp (kc Cj”=, (IIn DJ+, / Dj> / $I) Equation 3

The TOOLKIT option in MULTIFLO can be used to perform this calculation (use of theTOOLKIT option is described in Appendix A). The heat transfer coefficient through thecompletion can now be expressed (based on tubing inside diameter) as:

*kch =camp Equation 4

Di In (DC/ Dci)

At the bottom of the well there is normally only one tubing, one casing and one thickness ofcement (and one annulus). However, nearer the top of the well there may be numerouscasings, annulus, different annular fills and cement thicknesses.

6.2.4 Internal Heat Transfer

The internal heat transfer coefficient, hi, assumes the existence of homogeneous two-phaseflow. This assumption is usually acceptable since the resistance of the internal film to heattransfer is low compared with the resistance of the casings and the external medium.

The expression used for estimating the film coefficient is the same as that recommended forsingle phase turbulent flow.

O.O23k,,,hi = x ReO.* x Py0.3 Equation 5

Di

The physical properties are derived on a phase volume weighted basis. The subscript ‘m’,denotes a mixture property. For example, the mixture viscosity is determined by the following:

l4ll = CL, 4 + P&l - A.,) Equation 6

(Since heat is usually lost from well fluids the exponent for the Prandtl number [Pr = C, JJ,,/ k Jis set to 0.3.)

6.2.5 External Medium

Heat transfer to the earth is essentially transient radial heat conduction, with the effectiveheat transfer coefficient varying with time. Whilst the analysis of the problem is relativelystraightforward, a simple explicit solution is not available. However, approximate asymptoticsolutions for ‘short’ and ‘long’ flow times can be derived.

The effective heat transfer coefficient, referred to the tubing inner diameter, is given by:


BP Multiphase Design Manual Section 6. Wellbead Temperatures

*K?hearth

=--- Equation 7

Dir(t)

where, for short flow times (t<lOO hours):

f(t) = [(I@-~.~ + 0.5 - 0.25(z/~r)~~~ + 0.1254 x 2/D,,

DC,*z=-----

cd

t = time (hours)

and for long flow times (t>lOO hours):

Equation 8

Equation 9

f(t) = In(DcO / @) - 0.29 Equation 10

MULTIFLO will allow the user to specify the flowing time used in the equations above and willtherefore roughly calculate the WHFT for short flowing times (ie. warm-up of the well after start-up). However, this method will not provide an accurate prediction of the WHFT during start-up(although a rough indication is given). Therefore, WELLTEMP should be used to calculate wellwarm-up and not MULTIFLO.

For flowing times of the order of 10,000 hours, the system can reasonably be assumed to beapproaching steady-state, and MULTIFLO should be used. For times between 500 (say) and10,000 hours the well will probably be approaching steady-state, and the ‘steady-state’ valuegiven by MULTIFLO will be a reasonable approximation.

6.2.6 Geotherm Gradient

The ‘undisturbed’ temperature of the surrounding soil will vary from the reservoir to thesea/surface. The soil temperature is important since it affects the rate of heat loss from thefluids (see Equation 12). However, the geothermal gradient is not usually known and it iscustomary to assume that the (unaffected) temperature varies linearly between the reservoir andthe seabed.

6.2.7 Overall Heat Transfer

Using the coefficients calculated above the overall heat transfer rate can now be determined:

1 1 1 1-=-+-++

U hi hcomp hemth

Equation 11

The heat transfer coefficients are based on the inside area. The heat loss from the fluid istherefore:


Section 6 Wellbead Tmperatum BP Multiphase Design Manual

Q = UAST Equation 12

Both U and the temperature difference vary up the tubing and therefore a calculation is bestsolved by computer simulation.



6.3 Calculation of wellbead Temperatures

6.3.1 Effect of Flowrate

Increasing the flowrate will increase the wellhead temperature. Qualitatively, as the flowrateincreases, the heat ‘lost’ in the well (Q) remains roughly constant:

Q= “~Tnl Equation 12

The area is fixed and the overall heat transfer coefficient is relatively insensitive to flowrate sinceit is mainly controlled by the heat transfer through the surrounding rock. The Tlm will increaseslightly since the temperature difference at the surface will increase. The following equation canbe used when studying the heat loss from the fluid only:

Q Q == MCp,(Ti,MCp,(Ti, -T,,J-T,,J Equation 13

Therefore, rearranging Equation 13 gives:

Equation 14

Since Q and C, can be taken as constant (or relatively insensitive to the flowrate), increasing M(the mass flowrate) will reduce the temperature loss in the tubing. Therefore, increasing themass flowrate will resulting in an increase in the WHFT.

Equation 14 gives only a rough indication of the effect of flowrate since it fails to account for theeffect of the geothermal gradient. At very low flowrates the well head temperatures tendstowards the sea temperature as the fluid temperature tends towards the geothermal gradient.

6.3.2 Effect of Water Cut

Increasing the water cut increases the specific heat and mixture density of the fluid resulting inan increase in the WHFT. With reference to Equation 13 used above, by assuming that theheat lost from the well (Q) and the mass flowrate (M) to be relatively constant, increasingwater cut will raise the specific heat (Cd which will reduce the temperature loss and thereforeresult in warmer WHFT.

The predicted influence of the water cut and the flowrate is shown in Figure 1. This is agraph of the steady-state WHFT. Flowrates are shown up to 20 mblpd, or up to about 30 ft/s.The lines are asymptotic, and will approach the reservoir temperature (220°F in this case) atvery high flowrates.

The temperature approaches the asymptotic value at low flowrates (5 mblpd or roughly 5 ft/s),and at flowrates above this the WHFT increases slowly.


Section 6 Wellbead Tmperatums BP Multiphase Design Manual

200

190

180

170

180

0 2 4 8 8 10 12 14 18 18 20

Liquid flowrate (mblpd)

Figure 1. Relationship between WHFT and liquid flowrate and water cut

6.3.3 Effect of Gas Oil Ratio (GOR)

The gas phase in an oil well typically provides a relatively small part of the overall mass of theoverall mass of the mixture. Therefore a higher GOR only makes a minor contribution to theoverall enthalpy of the system. Since the rate of heat loss is dominated by the casings and theexternal medium, the increased velocity in the tubing has little effect and the heat loss remainsreasonably constant. The effect of the gas oil ratio is less significant than the water cut or theflowrate.

MULTIFLO was used to predict the WHFT for a range of GORs from 100 to 2,000 scf/stb (withthe same example as used previously). For a given flowrate, the WHFf only increased byroughly 2°F (1 “C) with increasing GOR. The range in WHFTs reduced with increasing flowrateand at 10 mblpd, the predicted WHFT only varied by 0.13”F between 100 and 2,000 scf/stb.

6.3.4 Effect of Lift Gas

210

n 70% water cut

0 50% water cut

A 30% water cut

A 20% water cut

0 10% water cut

0 0% water cut

Lift gas greatly complicates the heat transfer model. To rigorously model would require coun-ter-flow calculations for the lift section.

For long well flowing times the overall heat transfer is dominated by the heat transfer throughthe earth. As a result the WHFT will be reasonably accurate without considering countercurrentflow; provided the mixture temperature of the gas lift gas and produced fluids is calculated. Inthis simplified approach the temperature profile takes place at the gas lift injection point.

The BP Multiphase program does not calculate the mixture temperature (it assumes gastemperature is the same as the production fluids). As a result MULTIFLO will overpredict WHFTfor a well with high gas lift to liquid ratios if the gas lift temperature is significantly lower than theproduced fluid temperature.



6.3.5 Reliance in Predicted Results

As mentioned above, the calculation of the WHFT has a major impact in the design of thetopsides equipment such as the separator and the heat exchangers.

BP have performed a number of brief studies comparing actual WHFTs against predictedWHFTs for gas wells (West Sole & Bruce) and oil wells (Gyda & Ula). In general, the predictedresults differed from the actual results by less than 5°F. In most cases, these calculations wereperformed using reasonably accurate measurements of flowrate, water cut, and bottom holeflowing pressure and temperature taken while testing the well.

In most cases, the WHFT will be calculated using sketchy data from the appraisal wells whichmay not be suitable for the whole development. Usually, the most unreliable piece of data isthe reservoir temperature and the Bottom Hole Flowing Temperature (BHFT). This not onlyaffects the initial enthalpy of the wells fluids but also affects the geothermal gradient which isusually assumed to be linear between the reservoir and the surface.

Care must be taken when predicting WHFTs for gas wells since, although MULTIFLO accountsfor expansion cooling in the tubing, it cannot account for the cooling between the reservoir andthe perforations of the tubing. This may lead to the prediction of a high WHFT.

The thermal conductivity of the surrounding rock also has a significant influence on the calcu-lated WHFT. Usually only limited data on the rock type is available and the thermal conductivityis often assumed to be constant. It is advised that when calculating WHFTs, the sensitivity toreservoir temperature, flowrate, water cut, rock thermal conductivity are determined to providea range of temperatures.

issue 2 - December 1994 I I



LEFT BLANK



-

Unsteady-State (WELLTEMP)

6.4.1 Time Dependency

Before start-up the rock surrounding the completion will be at (or close to) the geothermalgradient. As the well starts to produce, the surrounding rock will start to warm up. Initially,there is a high AT between the fluid and the surroundings and heat loss is high. With time, theheat loss will reduce as the rock warms up and the WHFT will increase. The rate of thisincrease is dependent on the flowrate.

Well start-up can be of particular importance. Before the well is started-up, the temperature ofthe fluids is at the geothermal gradient. On start-up, ‘cold’ fluid will be produced, if the fluidspass through a choke at the wellhead then the temperature downstream of the choke may bevery low since the initial temperature upstream was low. This may result in hydrate formationwith the potential for forming a blockage.

It is therefore important to know how quickly the well would warm up in order to determine thelength of time that methanol/hydrate inhibitor injection would be required.

6.4.2 WELLTEMP

WELLTEMP is an unsteady-state single or two-phase simulation program. It can model manyunsteady-state temperature aspects of wells such as drilling and mud circulation. It’s primaryuse within the Multiphase Group of BP Exploration’s Technical Provision (XTP) is the modellingof the start up of wells.

The program requires fairly detailed data on the completion, such as tubing and casing diame-ters, annular fills, cement details and profile (measured depth and true vertical depth). Detailedcomposition cannot be input and the program uses basic oil data (such as density, viscosity,and yield point) with compositional data for the light end.

Capabilities of WELLTEMP include:

l Offshore wellsl Deviated wellsl Multiple geothermal gradientsl Productionl Injection

Direct comparisons between MULTIFLO and WELLTEMP have been favourable. It is possibleto approximate steady-state with WELLTEMP by specifying lengthy flowing times. UsuallyWELLTEMP predicts a WHFT l-2°F cooler than MULTIFLO. This could be due to either thesimulation of slightly different cases or that the WELLTEMP simulation has not come tocomplete steady-state.

Major differences in the WHTP prediction of MULTIFLO and WELLTEMP have occurred whenthere is a large section of the well tubing/casing surrounded by sea water. In this section therate of heat loss is high due to the sea temperature being low and the high heat transfer coeffi-cient between the outside of the casing and the water. Small difference in the specification ofthe problem may lead to the prediction of different heat transfer characteristics and therefore

issue 2 - December 1994 I3

Section 6 Wellbead Temperaturns BP Multiphase Design Manual

different WHFTs. This difference between the results of the two programs is more pronouncedwith greater riser heights.

It is therefore important that this section of the well is accurately specified.

6.4.3 start-up

When production commences from a well the WHFT will rise quite quickly. Usually, the wellhead temperature will approach the steady-state value within two residence times of the well.Typically this is no more than a few hours. The relationship between the flowing wellheadtemperature and the flowrate during the start-up as shown in Figure 2.

180

180

Ge 140 -

:ii 120 -8

t 100 -g!

P.- 80 -gE

g 8o n 20mmscfdf - 1Ommscfd3 40 A

0 5 mmscfd20 - + 2 mmscfd

0 I I I I I I0 1 2 3 4 5 6

Time (hours)

Figure 2. Well warm up rates, 4.5” tubing



Appendix: Setup of Well Using MULTIFLO

~A.I Introduction

This appendix of the Wellhead Temperatures section of BP Multiphase Design Manualdescribes the method by which a well is modelled within Multiflow. It is not intended to be aexhaustive description of the method (please refer to the MULTIFLO Manual) but it doesdescribe some problems involved in the calculation.

6~.2 Configuration

An oil or gas well can have many configurations (e.g. platform drilled, pre-drilled, subseatiebacks etc). Most wells can be suitably described by MULTIFLO by combining the Tubing,Risers and/or Pipes segments.

For an offshore well, from the reservoir to the surface the well must be described as Tubing.From the seabed surface, the well can no longer be described as Tubing since the Tubingsegments will not allow the external medium to be either air or water, therefore the well must bedescribed as a Riser.

If the system is a subsea tieback, then Tubing should be specified from the reservoir to thewellhead, and then Pipes and Risers in the normal way back to the platform.

The production fluids enter the well through the section of the tubing that has been perforated.This section cannot be modelled using MULTIFLO, and so the ‘inlet’ or the beginning of thesystem is usually at the top of the perforated section or the bottom of the smallest casing.

6~.3 Tubing

MULTIFLO models the tubing as having only one casing. In practice, there can be a number ofcasings and/or conductors. MULTIFLO models the remaining casings, annuli and cement asequivalent thickness of cement (the calculation of the equivalent thickness of cement isdescribed in Section A.7). The thickness of cement is entered in the Casing option of the Editmenu (shown below).

MULTIFLO Editor Dataset 53 (File 5) Forth Low Angle Well

Well Casing dimensions (in) Cement Heat

Segment Int diameter Roughness Thickness Thickness Xfer

1 8.688.68

22 8.688.68

33 8.688.68

44 8.688.68

0.00180.0018 0.470.47 1.3131.313 Y0.00180.0018 0.470.47 2.4562.456 Y0.00180.0018 0.470.47 3.6433.643 Y

0.00180.0018 . . . etc



6~.4 Annulus Mediums

The annulus between the tubing and the casing can be filled with either Oil, Water, Cement, orStagnant Gas.

Normally the annulus will be full of completion brine (i.e. Water). However, if gas lift is to beemployed, then the annulus will be used to convey the gas to the gas injection point. The gasprovides an insulating layer between the production tubing and the inner casing. The availableoptions do not cover this explicitly, however the stagnant gas and brine options should roughlybracket the GLT case.

6~5 Geothermal Temperature Profiles

It is unusual for a detailed geothermal gradient to be supplied with the completion data.Usually, the reservoir temperature and the sea temperature are supplied and a linear tempera-ture profile is assumed. MULTIFLO will allow the ‘top’ and ‘bottom’ hole external mediumtemperature to be added and will calculate the intermediate temperatures.

In the example below (within the HEAT option), the inlet to the first section and the outlet of thelast section have been entered.

What? Inlet Outlet

TUBING CALCULATED ROCK 1 4 0

TUBING CALCULATED ROCK





TUBING CALCULATED ROCK -47.6-47.6

By exiting the screen (pressing FlO or Gold-E) MULTIFLO will fill in the ‘missing’ temperatures.

What? Inlet Outlet

1 TUBING CALCULATED ROCK 140140 136.841136.841

22 TUBING CALCULATED ROCK 136.841 112.024



5 TUBING CALCULATED ROCK 85.4298 -83.819

66 TUBING CALCULATED ROCK -83.819 54.403254.4032

77 TUBING CALCULATED ROCK 54.403254.4032 - 4 7 . 6- 4 7 . 6

The temperatures now correspond to a linear geothermal gradient. The inlet and outlet temper-atures are stored with the information for that segment, changing the co-ordinates of thesystem will not automatically change the inlet/outlet temperatures to the geothermal gradient.The temperatures have to be refreshed by repeating the procedure described above.

Rock thermal conductivities and diffusivities are given in Appendix G of the MULTIFLO Manual.



6A.6 Pressure Drop Correlations

6A.6.1 Crude Oil/Gas Wells

It is important when sizing well tubing to use a correlation which is appropriate for the particularsystem. Numerous validation exercises have been performed, but no one correlation hasproved acceptable for all systems.

A recent extensive review of mechanistic models (performed by BP) has concluded that theAnsari method gave reasonable results over a wide range of conditions in oil and gas conden-sate wells, performing worst in high GLR conditions. In the absence of validation data for thefield in question, Ansari is the recommended correlation for vertical pressure drop.

6A.6.2 Gas Condensate Wells

The Gray correlation is usually preferred. However, Gray will give inaccurate results if flow isother than single phase gas, annular or annular mist. Hence if modelling a system with a highcondensate loading, or high produced water make, Gray should be avoided. The Ansarimethod generally gives good results over a range of possible liquid loadings.

GA.7 The TOOLKIT Option - Equivalent Thickness

6~.7. i Introduction

The TOOLKIT collection of utility programs includes COATINGS. This option allows the effectsof multilayer coatings on heat transfer to be calculated as an equivalent single layer.

6~.7.2 Example Problem Set-up

The text below follows a worked example.

The diagram below shows a tubing with three casing strings. The equivalent thicknessshould only be calculated for the layers outside the first casing. There is a water basedmud in one annulus and an oil based mud in the other, outside the third annulus there is a 2”thick layer of cement

Layer i-d. Conductivity Description

(inches) (Btu/hr/ft/OF

1 9.625

2 12.415

3 13.375

4 16.5

5 18

o.d. 24

0.36 Water Based Mud

26 Casing #l

0.141 Oil Based Mud

26 Casing #2

0.5 Cement

Thermal conductivity of equivalent layer = 0.5 Btu/hr/ft/“F

issue2 - December 1994 17


Equivalent thickness should only be calculated foroutside this first casing. Casing #l, Annulusand Tubing are not part of this calculation

T A CU N AB N SI u IN L NG U G

S#1

CASI

NG

#2

CASI

NG

#3

Layer 1 2 3 4 5

A.7.3 Coatings

The coating calculation option can be accessed through the TOOLKIT utilities menu by simplytyping COATINGS. The first screen appears as below:

MULTI-LAYER COATING DATA

Number of layers of coating EXCLUDING the wall/casing [

in the range [1:20]51

-.

In this case there are 5 layers outside the 9.625” casing. The first layer is the water based mudwith a thermal conductivity of 0.36 Btu/hr/WF. The data for the first layer can then be entered.

Collecting Data for Layer [ 1 1Inner Diameter for Layer [ 9.625 ] in inches

Thermal Conductivity for Layer [ 0.36 ] in Btu/hr/ft/OF

On pressing RETURN the inner diameter and the thermal conductivity of the first layer disap-pears and the program asks for the information for the second layer.

Collecting Data for Layer [ 2 1Inner Diameter for Layer [ 12.415 ] in inches

Thermal Conductivity for Layer [ 26 1 in Btu/hr/ft/OF

I 8 Issue 2 - December 1994

BP Multiphase Design Manual Section 6 Wellbead Temperatures

This is repeated until the inner diameters and thermal conductivities for all five layers have beenentered. The program now needs information on the o.d. of the outer layer and the thermalconductivity of the layer (usually cement) to which the multilayer system is to be equivalent.

Outer Diameter of Outermost Layer [ 24 1 in INCHES

Equivalent Conductivity of Single layer [ 0.5 ] in Btu/hr/ft/F

Inner Diameter of Flowline/Casing [ 6.184 ] in INCHES

The ‘Inner diameter of Flowline/Casing’isNOTpartofthecafculationbutisreported in the results summary. All the information has now been entered and the programcan now perform the calculation.

Flowline/Casing ID (in): 6.184

Flowline/Casing OD (in): 9.625

Equivalent OD (in): 38.60

Equivalent thick (in): 14.488

Flowline/Casing cond (Btu/hr/ft/OF): 26.00

Coating cond (Btu/hr/ft/OF): 0.50

To accurately model overall heat transfer, the equivalent OD should be reasonably accurate.Coating conductivity should be selected to provide correct OD.

The 5 layers have an equivalent thickness of 14.488” of a coating with a thermal conductivity of0.5 Btu/hr/ftPF. This can now be entered in the CASINGS option of the EDIT menu (seeSection A.3).

Moving up the well, the annulus material changes and the number of casings increase. Thecalculation must be repeated whenever the annulus medium changes and/or another layer isadded. A typical well may require 7-10 of these calculations to be performed. (It is advisablethat calculations begin at the bottom of the well since the program stores information from theprevious calculation and much of the information will for the subsequent calculation will be thesame or similar).

A.7.4 Riser Section

As mentioned previously, the TUBING option in MULTIFLO will not allow air or water to bethe external medium. Therefore these sections are described as RISERS.

Usually, if the well is a pre drilled well then the annuli will be filled with completion brine. Theequivalent thickness of ‘insulation’ can be approximated to half of the difference betweenthe o.d. of the outermost casing less the o.d. of the tubing. Small convection currents areset up within the various annuli containing the water, effectively increasing the thermal conduc-tivity of the water. This has been estimated to increase the thermal conductivity by a factor offive to 1.8 Btu/hr/W’F. (0.36 x 5).

The increase in the thermal conductivity due to the convection currents only applies to thecompletion brine in the riser section where the temperature difference is the highest. The oiland the water based mud are more viscous and convection currents are not thought to be


Section 6 Wellhead Tmperatum BP Multiphase Design Manual

generated. Hence it is recommended that the thermal conductivities for the oil and waterbased muds quoted in Section A.8 are used.

The conductivity of the steel is high and so it’s resistance to heat transfer can effectively beignored. The calculated equivalent thickness of cement must be entered as a SPECIFIEDlayer in the sub menu of the HEAT transfer section shown below:Segment Coating Characteristics

Standard Coatings USER SPECIFIED KP-ENTER to change

Thickness 7.25 (ins)

K Actual 1.8 (Btu/ft/hr/OF)

K Effective N/A [??? means pipe diam not set]

If the annuli of the riser section consists only of either of these two types of mud then the thick-ness of the ‘insulation’ can be calculated as above with the thermal conductivity of the puremud.

If the annuli are a combination of oil based mud, water based mud and/or completion brine,then a similar procedure will have to be adopted to calculate the equivalent thickness describedin Section A.7.3 above. The section from the seabed to the platform has to be modelled as aRISER to allow specification of air and sea as the eternal medium. The calculation of the equiv-alent thickness of cement must now include the annulus and the first casing i.e. the problemmust be specified from the outside of the tubing.

6A.8 Typical Thermal Conductivity Values

6~8.1 casings

The thermal conductivity of the casings and the tubing is high compared with the othermediums. Therefore the calculation of the equivalent thickness of cement is insensitive to thisvalue. For the range of temperatures normally experienced with typical wells the thermalconductivity of steel is 26 Btu/hr/WF.

-.

6~8.2 Cement

The thermal conductivity of a typical cement is 0.5 Btu/hr/WF.

6~8.3 Oil Based Mud

The thermal conductivity of the mud varies with density, different sources quote different values.a typical range is given below:


BP Multiphase Design Manual Section 6. Wellbead Temperatures

Density(w&)

Thermal Conductivity(Btu/hr/fV”F)

1.2 0.2002.0 0.343

1.08 0.2201.26 0.2451.51 0.285

6~8.4 H,O Based Mud

As with oil based muds, the thermal conductivity varies with density. Typical values are givenbelow:

Density(s4J

Thermal Conductivity(Btu/hr/ft/“F)

1 .o 0.361.2 0.3932.0 0.489




LEFT BLANK



A

%-Di

Do

DCI

Da3

hcowh earthhi

f(t)

kc

ke

k

M

TinToutQ

Nomendature

Area available for heat transfer (ft2)Fluid mixture specific heat (Btu/lb/F)Inside diameter (ft)Outside diameter (ft)Diameter to inside of concrete/coating (ft)Diameter to outside of concrete (ft)Completion heat transfer coefficient (Btu/hr2/ft/“F)External medium transfer coefficient (Btu/hr2/ftIoF)Internal film heat transfer coefficient (Btu/hr2/ft/“F)Function of time, Equation 8 (-)Cement thermal conductivity (Btu/hr/ft/“F)Soil thermal conductivity (Btu/hr/WF)Layer thermal conductivity (Btu/hr/WF)Fluid mixture thermal conductivity (Btu/hr/ft/“F)Mass flowrate (Ib/hr)Inlet temperature (“F)Outlet temperature (“F)Heat duty (Btu/hr)Dimensionless term in Equations 8 and 9 (-)Thermal diffusivity (ft2/hr)Temperature difference (“F)Log mean Temperature difference (“F)Non-slip hold-up (-)Vapour viscosity ([ft2/hr)Fluid mixture viscosity (ft2/hr)Liquid viscosity (ft2/hr)Fluid mixture density (lb/ft3)Reynolds NumberPrandtl NumberOverall heat transfer coefficient (Btu/hr2/W’F)


Section 6 Wellhead Tenzperatuves BP Multiphase Design Manual

--


LEFT BLANK


Section 7. Transient Flow

Section 7 Transient Flow

7.1

7.2

7i3

7.4

7.5

7.6

Introduction to Transient Multiphase Flows

Transient Programs in Use

Transients due to Flowrate Changes

Transients due to Topographical Effects

Transients due to Pigging Pipelines

Pressure Surge Transients in Two-phase Flow

Examples


Section 7. Transient Flow BP Multiphase Design Munual


LEFT BLANK


7.1

--

Introduction

Steady state operation is the exception rather than the rule in multiphase systems due to anumber of factors such as the nature of the flow regimes, interactions with the flowline profile,and changes in supply or demand. As a result several types and speed of transient need to beconsidered in order to size separation equipment, to determine operational procedures toachieve maximum cumulative production, to ensure safe operation, and to assess the impact ofpipeline ruptures, for example.

This section of the Multiphase Design Manual aims to describe where transient multiphaseflows occur and what design tools are available to the engineer. A number of sections aredevoted to the different types of transients which have been categorised as follows:

l Transients caused by flowrate changes

l Transients resulting from pressure changes

l Transients created by topographical effects

l Pigging transients

l Pressure surge transients

Transients caused by pipeline ruptures are covered separately in Section 9 of this DesignManual.

Over the past few years as dynamic simulators have been developed, their use has becomewidespread, and central to the determination of efficient operating guidelines. This has beendriven by costly lost production due to unforeseen transient flows causing plant shutdowns(such problems may not have been expected from steady state analysis), and the increasedconcern for the safety and environmental impacts of pipeline failures, which often results in tran-sient two-phase flow, even though the normal pipeline operation may be single-phase. As tran-sient simulations have become available they have also found use in other aspects of pipelineoperation including corrosion assessment and pipeline leak detection.

In parallel with computer code developments has been the increasing complexity of multiphaseproduction systems. Such schemes have generated the need for greater sophistication in mul-tiphase simulation and have been part of the driving force behind the development of the codesavailable today. In addition the rapid advance in computer technology has meant that expen-sive and time consuming calculations that were previously run on large company main framemachines are now efficiently run on desk top machines and even portable personal computers.This has greatly improved the access to transient simulators and significantly reduced the costof running the codes.

This chapter of the manual is designed to show the reader the types of transient two-phaseflow simulations that can be carried out by current technology. The aim is not to swamp thereader with the mathematics of the codes but to show by examples some of the work under-taken in the past. Simple hand methods are outlined where possible, however the complexityof transient two-phase flow analysis usually requires computer solutions provided by specialistcodes.

-.


Section 7. Transient Flow BP Multiphase Design Manual


LEFT BLANK


7.2 Transient Codes in Use

Although there are a large number of engineers and scientists working in the area of transienttwo-phase flow modelling, the number of commercial software packages is limited to just a few.The most widely used is the OLGA code developed by IFE and SINTEF in Norway. OLGA wasthe first commercial general transient two-phase flow package, which has enjoyed wide use bythe 6 members of the consortium that sponsored the development. The early versions of thecode required considerable expertise to create input data and to interpret the results. As aresult many of the oil companies using the code found it difficult to use and very costly. Someof them have developed their own versions of OLGA for their specific needs.

In competition to OLGA is the PLAC code developed by Harwell laboratories which was onlycommercialised in the autumn of 1992, but had been used for consultancy for more than twoyears previous. PLAC was developed with support from BP and is the transient simulator mostwidely used in-house by BPX.

There are several other transient two-phase flow codes presently under development. Total areworking on a transient version of the mechanistic PEPITE program called TACITE, which willfeature a slug tracking model. Shell have a code called TRAFLOW which is under developmentas an on-line simulator and analysis code.

There is a great deal of activity in the development of transient programs. However most of thisis being carried out in-house by oil companies and research centres, hence little is known aboutthem. For this reason the discussion in this Section will be limited to the codes that BPX havehad direct experience with. It is however expected that competition in the market place willincrease when some of the other codes are commercialised.

A description of the BP developed BPREAKL code for estimating release rates from volatile oilpipelines is given in Section 9 of the Multiphase Design Manual.

OLGA

OLGA is a dynamic, one dimensional modified two fluid model for transient two-phase hydro-carbon flow in pipelines and networks, in which processing equipment can be included. Thecode has been developed by the SINTEF/IFE two-phase flow project which commenced in1984 and was based on the computer program OLGA 83. This was originally developed forStatoil by IFE in 1983.

The two-phase flow project was aimed at improving OLGA by expanding an experimental data-base from a high pressure 8” large scale test facility run by SINTEF at Tiller in Norway.Extensive testing was carried out by IFE and by the projects member oil companies whichincluded Conoco Norway, Esso Norge, Mobil Exploration Norway, Norsk Hydro, Petro-Canada,Saga Petroleum, Statoil, and Texaco Exploration Norway.

The basic models in OLGA contain three separate mass conservation equations for the gas, thecontinuous liquid and the liquid droplets, these are coupled with the interphase mass transferterms. Momentum conservation is applied to the gas-droplet field and to the continuous liquid,hence giving two-equations. The mixture energy equation is written in conservation form,accounting for total energy balance in the system. OLGA predicts as a function of time thepressure, temperature, mass flow of gas and liquid, the holdup and the flow pattern. Closurelaws are required for the friction factor and the wetted perimeters of the phases and these are


Section 7. Transient Flow BP Multiphase Design Munuul

flow regime dependant. The droplet field also requires a entrainment and deposition model.The two basic flow regimes adopted are distributed and separated flow. The former containsbubble and slug flow and the latter stratified and annular flow. The transition between the tworegime classes are determined according to a minimum slip concept.

Due to the numerical solution scheme the original versions of OLGA are particularly well suitedto simulate slow mass flow transients. The implicit time integration applied allows for long timesteps which is important for the simulation of very long transport lines, where typical simulationtimes are in the range of hours to days.

The necessary fluid properties (gas/liquid mass fraction, densities, viscosities, enthalpies etc.)are assumed to be functions of temperature and pressure only, and have to be supplied by theuser as tables in a specific input file. Thus, the total composition of the two-phase mixture isassumed to be constant both in time and along the pipeline for a given branch. The user mayspecify a different fluid property table for each branch, but has to ensure realistic fluid composi-tions if several branches merge into one.

In 1989 Scandpower A/S acquired marketing rights to the OLGA program and commercial useof the code increased considerably. The early versions of OLGA had a simple data input andoutput file format which made the setting-up of problems very time consuming. Later versionshave user friendly interfaces which greatly enhances the use of the code, PreOLGA is used forinput data generation and PostOLGA is used for post processing. Fluid properties are generat-ed using a package called PVTOL.

In 1993 Statoil acquired the Tiller facility for 5 years and embarked on a substantial R and Deffort to improve OLGA, areas under development include:

l Improved mechanistic modelling and closure laws

l Slug flow modelling

l Three phase flows

l Compositional tracking

These areas also indicate where there are deficiencies in the present code.

In addition, OLGA has been interfaced with the Fantoft D-SPICE multi-compositional dynamicprocess plant simulator to improve the modelling of equipment, however this interface only pro-vides a dynamic exchange of data between the codes and is hence not a fully integratedmodel.

Some of the other original two-phase flow project sponsors have also developed OLGA for theirown use, notably Conoco who have developed improved thermal modelling and pigging simu-lation in their version called CONOLGA.

PLAC

The PipeLine Analysis Code (PLAC) was originally developed from the nuclear reactor safetycode TRAC (1986) under a four year programme sponsored by BP Exploration and the UKOffshore Supplies Office. The major parts of that work were to remove the redundant reactorspecific components and to convert from steam/water physical properties to multi-component



-

hydrocarbon properties. More suitable models for interfacial friction and flow regimes were alsointroduced. The first release of PLAC was available for use by AEA Petroleum Services and BPEngineering in mid 1990. Commercialisation of the code was not realised until the autumn of1992 however BP and AEA conducted numerous validation tests on the code with good andbad experiences. PLAC is the transient simulator used most widely in-house by XFE to modeltransient two-phase pipeline flows. Like OLGA the original versions of the code were cumber-some to use and the ease of use has been greatly improved by the development of interactivepre and post processors.

PLAC solves mass, momentum and energy equations for each phase using a one-dimensionalfinite difference scheme. Six equations are solved, gas and liquid mass and momentum con-servation, total energy conservation, and gas energy conservation. Unlike OLGA, PLAC doesnot have a separate equation incorporating a droplet field in the gas stream. Appropriate flowpattern maps and constitutive relationships are provided for wall and inter-facial friction and heattransfer, and a model for multi-component phase change is included. PLAC has flow regimemaps for vertical and horizontal pipes and switches to vertical flow if the angle of inclination isabove 10 degrees. The horizontal flow pattern map is based on the method of Taitel and Dukler(1976).

The fluid physical properties are calculated from a user supplied mixture composition using aninternal PVT package, however this generates a table of properties similar to OLGA, and hencestill relies on simple equilibrium phase behaviour predictions. PLAC can only use one composi-tion in a network. The current version of the code has the capability to handle PIPES, TEESand VALVES and is able to predict a range of phenomena including flowrate and/or thermaltransients, severe slugging, shutdown/restart problems and pipeline depressurisation. Theboundary conditions are specified by FILL and BREAK components.

TRA.NFLO

TRANFLO is a three phase flow transient program developed by the mathematical sciencesgroup at BP Research, Sunbury. The program is based on the one-dimensional averagedequations of motion for the conservation of mass and momentum for each phase and does notinclude the effects of gas entrainment. The equations for each phase are transformed into evo-lution and algebraic equations. The resulting system of equations is analogous to that formotion of liquid down a open, inclined channel. The program can generate waves on the inter-face due to physical instabilities. These evolve non-linearly into a train of periodic waves whosefronts propagate as shocks or hydraulic jumps. The waves can touch the top of the pipe lead-ing to the formation of slug flow.

TRANFLO was under development at Sunbury on the VAX computer and showed somepromise in the simulation of sloshing in dips, slug tracking, and water puddling in pipelines. Theprogram development ceased in 1993 and was handed over to AEA Petroleum Services,Harwell, where its future is unclear, however it may be incorporated into PLAC or ported onto aPC. The version at BP is limited to isothermal cases without phase change, and is also poor forvertical flow. It is an R&D tool and is not suitable in its present form for modelling completepipeline systems, but may be useful for investigating three phase effects in simple specificgeometries. The three phase version of PLAC is expected to make TRANFLO obsolete in thefuture.


iSection 7. Transient Flow BP Multiphase Design Munud

SPEEDUPSPEEDUP is a general purpose simulation package capable of solving coupled, ordinary differ-ential equations. It is used within BP to examine transient conditions in process facilities suchas advanced control of reactors and distillation columns, the interaction of slugging oil and gasproduction lines with facilities, compressor control and surge, stream fuel gas and relief sys-tems etc..

SPEEDUP is also capable of addressing a wider range of problems. It can perform steady statecalculations, dynamic simulation, optimisation, parameter estimation and data reconciliation. Itis also a flowsheeting package and data is organised in terms of streams of connecting units.The units in turn are models taken from commercial or user generated libraries.

The program is generally used by the dynamic simulation group to investigate the response ofprocess plant to multiphase pipeline transients such as severe slugging and normal slugging.This is handled by a Dynamic Slugging Model that has been incorporated as a module withinSPEEDUP and allows and integrated model of the pipeline and facilities to be used to assessthe effect of slugging on the entire process plant. The model simulates the deceleration of theslug as it enters the riser (due to the increase in the hydrostatic head) and the subsequentacceleration of the slug as it exits the riser and enters the topsides facilities. The slug produc-tion phase is followed by a period of gas blowdown as the pressured gas bubble behind theslug depressures into the facilities. The Dynamic Slugging Model assumes a two-phase system(liquid and gas) and requires the size and frequency of the slugs to be input. It does not allowslugs to break-up or to amalgamate.

WELLTEMP

WELLTEMP was designed to examine temperature transients caused by flow in wells. It canalso be used to examine temperatures in pipelines. It handles heat flow through tubing, cas-ings, annuli, cement and formation. various operations can be modelled such as injection, pro-duction, forward and reverse circulation, shut-in and cementing. It assumes steady stateflowrates; although different flow periods can be entered (steady within each period). It can han-dle dual or even triple completions (and co-bundled flowlines). In dual completions it wouldallow injection down one string and production back up the other.

It can be used to examine surface and downhole tubing temperatures during drilling or simulat-ed treatments. This is useful when considering fluid properties, cement characteristics etc. Itcan also help when considering corrosion, wax deposition or hydrate formation. It can be usedto examine cool down times in pipelines. It could also be used to refine the temperature profilein the annulus and wellbore during gas lift operations.

WELLTEMP is marketed by Enertech Engineering and Research Co of Houston and has beenprimarily used by XFE to investigate the warm-up of wells and flowlines. WELLTEMP can usu-ally run much quicker than PIAC and has been used to investigate sensitivities before perform-ing a detailed PLAC simulation. The simple composition treatment in WELLTEMP is a limitationto detailed transient two-phase flow analysis.

University of Tulsa Codes

The University of Tulsa through the TUFFP JIP have developed a number of transient codesover the years which have been made available to BP. One of the earliest codes was produced


by M W Scoggins in 1977 and was based on a one-dimensional two-phase flow model withslip. The model is based on the existing steady state Eaton method for holdup and the Duklertwo-phase friction model. The treatment of phase behaviour and physical properties is basedon hard wired black oil correlations and the simulation is isothermal only. This program hasbeen ported onto a PC by XFE and is called SCOGGINS, however the code suffers fromnumerical problems and its inherent limitations mean that it is little used.

The latest transient analysis code from TUFFP was developed by Minami in 1991 who imple-mented a simplified model proposed by Taitel. He also developed a new pigging model thatwas coupled to the transient model. The sets of equations are discretized using a lagrangeangrid system for the pigging model. The resulting set of equations are solved with a semi-implicitscheme and an approach based on the stability of the slug flow structure is used to predict theflow pattern transition boundaries.

The version of this code released to BP is only capable of simulating conditions in the TUFFP1400 feet long 3” test facility and is hence of little use. Comparison of the predictions withrecent data from transient tests on the Tulsa facility indicate that the code is inferior to PLAC.

FLOWMA!WER

Flowmaster is a computer program designed primarily for single phase fluid flow analysis andpipe network design, and has various analysis options including steady state hydraulic analysis,flow balancing, and pressure surge analysis. The network is constructed from a pallet of com-ponents and the design data input for each item. XFE primarily use FLOWMASTER for singlephase pressure surge analysis, however there are a number of features which can facilitate asimple surge analysis for two-phase flows including a bubbly pipe component, which allows thespeed of sound to be modified to account for a homogeneous gas fraction, and a primer com-ponent, which can simulate liquid filling a gas space.




LEFT BLANK



7.3 Transients due to flowrate changes

The operation of oil and gas pipelines often involves changing the inlet or outlet flowrates for anumber of reasons, namely:

l Start-up and shut-down.

l Fluctuating supply and demand, i.e gas delivery changes.

l Switching wells on and off for maintenance or testing.

Such flowrate changes can have a considerable effect on the downstream processing plant asflowrate increases can give rise to high transient liquid production rates and potential gassurges, and reductions in flow can result in periods of low flow. This is illustrated by the follow-ing example.

7000 -

6000 -

5000 -

4000 -

3000 -

2000 -

1000 -

o-t-0

- Minami-1

BBR

Eaton

- MeasuredDukler

No-Slip

I I I I I

2 4 6 8 10

Gas flow rate (mmsm3D)

Figure 7.1 Marlin pipeline holdup profile

Figure 7.1 shows the predicted liquid content of the Marlin gas condensate pipeline as a func-tion of the gas flowrate. This is a 67 mile long, 20” diameter wet gas line operating with a liquidloading of around 65 bbls/mmscf. It is seen that the general trend is that the liquid contentreduces as the gas flowrate increases. Hence, if a gas flowrate increase is made, liquid will beremoved from the pipeline as the new equilibrium liquid content is established. If the change inthe gas flowrate is carried-out too fast, the excess liquid can be swept out as a large ‘slug’which may overfill the downstream plant, whereas if the changes are made slowly, the liquidcan be gradually swept out within the slug catcher capacity. It is also interesting to note in thisFigure the wide variation in the predicted holdup obtained from using the various methods.This may not be too much of a problem when calculating the amount of liquid swept out duringa transient because it is the difference in the holdup that it most important. In this example theEaton holdup correlation is expected to give reasonable answers. However, one should bewary of using the correlation approach rather than mechanistic models since they do not usual-ly predict the steep rise in holdup as the velocity and inter-facial friction reduces - hence a grossunderestimate can result when starting from low flowrates.

Issue 2 - December 1994 II

Section 7. Transient Flow BY Multiphase LIesign Ma?Uhd

Figure 7.2 shows the liquid flowrate at the outlet of the Marlin gas condensate pipeline duringan increase in the flowrate from 155 mmscfd to 258 mmscfd. During the test the gas rate washeld constant at 155 mmscfd for 52 hours in order to reach equilibrium conditions. The ratewas then increased to 258 mmscfd in a period of one hour and held constant for a further 26hours to obtain equilibrium conditions again. It is seen that during the transient the outlet liquidflowrate is considerably higher than the final equilibrium value.

Marlin line rate change test

- - - - - - - - - Predic tedf low- Measured flow

3pt. moving avg.

1,200 - ------ Measured flowsmooth curve

1,000 -

Transition period gas rate = 258 MMscfdCondensate rates:Predicted =742 + 3,343 bbl / 21.4 hr

= 898 bbl / hrMeasured = 895 bbl / hr

800 -

Condensateflow to slugcatchers,

bbllhrFinal gas rate = 258 MMscfdCondensate rates:Predicted = 742 bbl/hrMeasured = 727 bbl/hr

200 -Condensate rates:Predicted = 69 bbl/MMscf x 6.46 MMscf / hr

= 445 bbl/hrMeasured = 424 bbl/hr

I I I I I I I I I I I I I I I I I I I I I I I I

1600 2000 2400 0400 0800 1200 1600 2000 2400 0400 0800 1200 1600

Oct. 12 Oct.13 Oct.14

Figure 7.2 Marlin rate change test

5000 11--.--- Predicted flow rate- Measured flow rate

4000

3000Outlet liquid

flow rate(M3/D)

Temperature = 13°CCondensate gas ratio = 3.6 x 10 -4 M3/M3Discharge pressure = 7.8 x 10 s N/M*

I5

I I I I I10 15 20 25 30

Time (Hours)

Figure 7.3 Marlin comparison with SCOGGINS

In the next section we shall outline a simple method for estimating the slug catcher sizerequired to handle the liquid removed during flowrate increases. Such a method was used todetermine the predicted flowrate profile shown in Figure 7.2. In later sections the use of tran-sient computer codes will be outlined. However, Figures 7.3 to 7.5 are provided here to illus-trate the potential accuracy of transient codes on the Marlin rate change data. This data hasoften been used as a test case as it is one of the few transient field data sources available in thethe open literature.



6ml6ml II // // II II ,,00 25 35 45 55 65 75

Time (hr)

Figure 7.4 Marlin comparison with PLAC

s 8 0 0

1 - - Measured flow, 3 pt moving average- Measured flow, smoothed curve1.1 OLGA

II I I I I I I I I I I I

1600 2400 0800 1600 2400 0800 1600Ott 12 act 13 act 14

Figure 7.5 Marlin comparison with OLGA

Pigging gas condensate pipelines can also result in large slugs. However, in many cases it isnot feasible to design slug catchers of a sufficient size to handle the equilibrium slug producedby running pigs at low throughputs. In this case pigs must be run frequently to prevent the liq-uid holdup reaching equilibrium, which can incur high operating costs. Sometimes gas ratescan be reduced during the pig arrival to allow the produced liquids to be processed, but in oth-ers the gas flowrate may be determined by the consumer, and the pigging operation carried outat the prevailing gas rate. In some situations pigs are only required on an infrequent basis forcorrosion control or inspection, in which case the pipeline liquid content may be high and a pro-cedure must be put in place to handle the liquid swept-out by the pig. One way of doing this isto stop the pig offshore and ‘walk it’ into the slug catcher at a rate compatible with the liquidprocessing capacity. This approach was successfully carried out at the restart of pigging oper-ations on the Amethyst pipeline where a liquid volume of over four times the slug catcher-


Section 7. Transient Flow RP Multiphase Design Manual

capacity was allowed to accumulate in the sealine when the offshore pig launcher failed. Insome instances it is possible to reduce the liquid content of the pipeline prior to pigging by con-trolled rate increases to remove liquid. In other cases pigging is not possible, and flowratechanges must be controlled to prevent overfilling the downstream plant.

With these factors in mind it is seen that for gas condensate systems at least, it is often thetransient ‘slug’ that determines the slug catcher volume. For oil and gas pipelines pigging maybe required frequently for wax control etc, where the lines have reached equilibrium, hence thismay determine the size of the slugcatcher. However for some developments, particularly sub-sea, pigging is required less frequently and can be accomplished with some operating ingenu-iry. Here it is often the longest normal slug or the transient rate change liquid sweep out thatdetermines the required slug catcher surge volume. The next section outlines a simple calcula-tion method for estimating the liquid outflow profile due to rate change transients.


Simple analysis methods for flowrate increases and reductions

The first step in considering the required surge volume for transient flowrate increases can bemade by considering the equilibrium holdup/flowrate profile.

25

15

Slug size(1000 BBLS)

10

5

0 7-

0I 1 1

100 200 300 400

Gas flowrate (MMSCFD)

Figure 7.6 Amethyst pipeline holdup profile

Figure 7.6 shows the equilibrium holdup for the Amethyst pipeline in the Southern North Sea.The line is 30” diameter and connects two unmanned production platforms with the Easingtongas processing terminal 30 miles away. The working volume of the slug catcher is 3800 bbls,with a maximum liquids pumpout rate of 6000 bbl/d. The liquid loading is typically 6.5bbl/mmscf of condensate and 0.54 bbl/mmscf of water and methanol. It can be seen that fornormal gas flowrates in the region of 250-350 mmscfd the equilibrium liquid holdup is well inexcess of the slugcatcher capacity, hence pigging at equilibrium conditions could overfill theprocess facilities. Higher gas rates than those available would be required to give equilibriumholdups below the surge volume. In fact this pipeline was designed to be pigged daily, andhence the liquid expected to be received was the order of 1760 bbls at 250 mmscf/d. Thisdesign allows for a missed pig, ie two days between sphere launches, without causing liquidhandling problems.

The holdup profile shows a rapid rise in the liquid content at gas rates below 200 mmscfd,hence operating in this region could cause problems even without pigging, as relatively smallgas rate perturbations can cause large quantities of liquid to be swept from the pipeline. Forexample a 10% increase in the gas rate from 200 to 220 mmscfd could remove 4750 bbls ofliquid and, possibly, swamp the slug catcher.

Shortly after the start-up of the Amethyst pipeline system the pig launcher failed and it wasdecided to investigate the consequences of continuing production, and allowing the pipeline liq-uids to build up to equilibrium values. A decision was taken to limit the minimum gas flowrate to250 mmscfd, and hence to avoid possible uncontrolled sweep-out due to inherent flowrate



fluctuations. The pipeline was operated without pigging over the winter, where it took severalmonths to obtain an equilibrium holdup, which was estimated to be in the region of 18000 bblsfrom a mass balance. This is within 17% of the value predicted by the old segregated flowmechanistic model in MULTIFLO, but is 213% higher than the value predicted by the Eaton cor-relation. Some of the simple transient analysis outlined below was used to investigate how toresume pigging operations.

A simple approach to sizing a vessel to handle the liquid produced by a gas rate increase wouldbe to consider the change in the equilibrium holdup and ignore the effect of the liquid pump-outrate. For example if the gas rate were increased from 250 mmscfd to 350 mmscfd the equilibri-um liquid removed would be 15373 - 8018 = 7355 bbls. Hence the gas rate could be increasedin two steps from 250 mmscfd to 300 mmscfd which would remove 3800 bbls and then 300mmscfd to 350 mmscfd which would remove 3555 bbls. A 7355 bbl surge volume would berequired if the increase were made in one step without taking account of the liquid pump-outrate. The maximum gas flowrate available is 350 mmscfd which gives a pipeline inventory of8018 bbls, and hence it is still necessary to ‘walk in’ the first pig.

We will use the Amethyst case to illustrate a simple way of determining the effect of the pump-out rate on the required surge volume. Consider the case of an increase in the gas rate from250 to 350 mmscfd.

At 250 mmscfd the equilibrium holdup is 15373 bbls

At 350 mmscfd the equilibrium holdup is 8018 bbls

Hence the difference in holdup is 15373 - 8018 = 7355 bbls.

Next calculate the initial and final liquid production rates from:

Liquid rate = liquid loading x gas flowrate

Initial liquid flowrate = 250 mmscfd x 7.04 bbl/mmscf = 1760 bbl/d

Final liquid flowrate = 350 mmscfd x 7.04 bbl/mmscf = 2464 bbl/d

Calculate the duration of the transition time for the transient, which is the length of time overwhich the high flowrate occurs. If it is assumed that all the liquid in the line accelerates to theequilibrium liquid velocity corresponding to the final gas rate, then the transition time is thesame as the residence time at the final rate, ie:

Transition time = (final holdup / final flowrate) = (8018 / 2464) = 3.25 days

The transition flowrate is the sum of the final flowrate and the increase due to the rate changeand is given by:

Transition flowrate = Final flowrate + (holdup change /transition time)

= 2464 + 7355/3.25 = 4727 bbl/d


It is seen that based on this method the surge can easily be handled by using the 6000 bbl/dpump-out capacity of the Easington terminal. Figure 7.7 shows the predicted liquid outflowprofile.

Figure 7.7. Simplified liquid outflow profile during ramp-up

Liquidoutflow

rate(BBUD)

-1 0 1 2 3 4 5Time (Days)

This approach can be extended to investigate the trade-off between the required slug catchersurge volume and the pump-out rate using the relation below:

Surge volume = transition time x (flowrate in - flowrate out)

= Tt x (a,, - Q,,J

If the pump-out rate is fixed at the final equilibrium value the required surge volume is:

V5 = 3.25 ( 4727 - 2464 ) = 7355 bbls

ie. the change in equilibrium holdup.

If the pump-out rate is 4727 bbl/d then this method shows that no surge volume is required.The solution to the equation is the linear relationship shown in Figure 7.8. It can be seen thatfor a surge volume of 3800 bbls a minimum pump-out rate of 3550 bbl/d is required.

The relationship shown in Figure 7.8 can be subject to large inaccuracies at the extremes asthe flowrate tends to zero and at the final transition flowrate, the reason being as follows; atzero pump-out rate the surge volume is equal to the transition flowrate multiplied by the transi-tion time, whereas in practice liquid continues to flow into the vessel at the final equilibriumflowrate, hence the required surge volume becomes infinite as the pump-out rate goes to zero.When the pump-out rate is equal to the transition flowrate the above method indicates that nosurge volume is required. However in practice the flowrate during the transient is not usuallyconstant, and typically peaks at the start. Hence the solution for a surge volume of zero is apumpout rate equal to the peak flowrate during the transient.



Surgevolume(BBLS)

16000 -

8000 -

6000 -

0 I I I I I I I I I I0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Liquid pump-out rate (BBLS/D)

Figure 7.8 Surge volume as a function of processing rate

The same simple method can also be applied to the estimation of the outlet flowrate profile dur-ing a flowrate decrease, as follows: Consider a reduction in the gas flowrate of the Amethystpipeline from 350 mmscfd to 250 mmscfd.

First calculate the residence time at the final flowrate, this is also assumed to be the transitiontime:

Transition time = (final holdup / final flowrate) = (15373 / 1760) = 8.73 days

then:

Transition flowrate = Final flowrate - (Holdup change / Transition time)

= 1760 - (7355 / 8.73) = 918 bbl/d

Hence the hand calculation method predicts an initial fiowrate of 2464 bbl/d falling to 918 bbl/dover a 8.73 day transition period after which the rate increases to 1760 bbl/d. This is illustratedin Figure 7.9.

Example 7.1 illustrates a PL4C simulation of the shutdown and restart of Pompano subseawells and Example 7.2 gives a comparison with the simple hand calculation method.


Liquidoutflow

rate(BBUD)

-

2500

2000-I

1500-

1000-

500

1

nI I I

-6 -4 -2 0 2 4 6 8 10 12

Time (Days)

Figure 7.9 Simplified liquid outflow profile during ramp-down


Section 7. Transient Flow BP Multil,huse Design Munual


LEFT BLANK


7.3.2 Use of transient computer codes for other boundary conditionchanges

There are other changes in the boundary conditions of a pipeline apart from flowrate changes,for example outlet pressure changes, which are often required to be simulated by transientanalysis. With general transient simulation programs such as PLAC there is a wide variety ofboundary conditions that can be changed. For example PLAC uses the FILL component tospecify a flowrate boundary and the BREAK component to specify a pressure boundary. Whenrunning PLAC the user can specify signal variables which are used to vary the conditions at theinlet or outlet and to control the action of valves. One of the simplest ways of changing aboundary condition is to select time as the signal variable, and to the define the flowrates at thefill as a function of time. In this way rate changes can be simply entered as a table of values.PlAC assumes a linear change of the variable with time. There are 25 signal variables availablein PLAC at present, allowing a wide range of boundary condition changes such as pressure,temperature, flowrate. heat transfer and flow area. It is beyond the scope of this chapter to con-sider all combinations, and hence reader is directed to Example 7.3 as an illustration of aboundary condition problem.

Thermal transients such as pipeline cool down and warm up are important to the estimation ofthe amounts of inhibitor that is required to prevent hydrate formation during periods of zero orlow flow. Here WELLTEMP can sometimes be used to screen cases for detailed analysis bymore sophisticated transient two-phase flow codes.

Pipeline pack and draw is also often modelled to investigate the survival time of a given systemand the ability to meet the gas sales contract.

-

issue 2 - December 1994 2 1

Section 7. Transient Flow BP Mztltiyhase Design Manual


LEFT BLANK

-


BP .M~~ltiphcrsc Ile.sicqli .Mmlrlrrll Section 7. Transient Flow

7.3.3 Hydrate problems due to shut-down and start-up

In normal operation hydrates can be controlled by maintaining the flowing temperature abovethe hydrate formation point and/or injecting sufficient quantities of inhibitor. The requireddosage is usually determined from running GENESIS, HYSIM or using the GPSA charts for theprevailing conditions. However, during shut-down and start-up, conditions can occur whichincrease the likelihood of hydrate formation, and hence require additional quantities of inhibitor.Figure 7.10 shows a typical hydrate dis-association curve. A pipeline pressure/temperatureprofile is superimposed as line ‘A’ representing conditions in a pipeline from a subsea templateto the slug catcher on a host platform, where hydrate inhibition may not be required under nor-mal operating circumstances. In some cases the pipeline may be long and the fluids may coolto seabed conditions where some inhibition may be required. This could also occur as a resultof lower flowrates, and is shown as curve ‘B’. Hence, in normal operation steady state analysiscan be used to determine the amount of inhibitor required.

100

90

80

70

Hydratesand ice

Ice

10%MeOH

Hydrates

/

I I0%

MeOH

Wellhead

-5 0 5 10 15 20

Hydrate dis-association temperature (“C)

Figure 7.10 Hydrate formation conditions

Separator


Section 7. Transient Flow BP Multiphase Des&n Manual

During start-up or shut-down transient conditions may arise which can lead to operation in thehydrate region. For example, as the line is started-up the pressure can rise quicker than thetemperature giving the pipeline profile drawn in line ‘C’. It can be difficult to ensure that therequired inhibitor dosage is present during such start-up transients, and hence it is common forinhibitor to be pumped into a pipeline before shut-down to protect against controlled start-upand shut-downs. During unforeseen transients special care may be required.

Another aspect which can complicate the determination of hydrate formation conditions is thetime required to form hydrates, ie. the kinetics of the process. If the conditions during start-upare in the hydrate formation region for only a short period of time, then extra inhibition may notbe required.


Transients Due to Topographical Effects

7.4.1 Hilly terrain slug flow and dip slug generation

In all the classical steady state programs used to model multiphase flows the flow regime isassumed to change abruptly as the topography changes. However, in practice the previousflow regime may persist for sometime in the next section. An example of this is flow over a hillwhere slug flow may be predicted in the uphill section and stratified in the downhill part. In asteady state simulation the slugs would abruptly disappear at the apex. In practice this may nothappen, and the liquid slugs may persist for sometime in the downward sloping section whichcan give rise to a much higher pressure recovery than in stratified flow. This phenomena hasbeen addressed in part by the spreadsheet program HillyFlo, which has been developed byXFE to model slug flow in hilly terrain pipelines. HillyFlo is a steady state program that deter-mines the slug size in uphill inclined sections and models the slug decay in downward sections,hence allowing for the pressure recovery.

The interactions with the terrain may cause additional effects. For example, the liquid sloshingin a dip may cause perturbations in the gas and liquid flowrates downstream, giving rise to flowregimes different to those expected from analysis with steady flowrates. Flowrate fluctuationscaused by slugs dissipated in a downward sloping section may be sufficient to generate slugsin a following downstream section normally operating in stratified flow.

By taking account of the hysteresis in the flow regimes and by tracking the local phase flow-rates it is easy to see how in practice flows may differ greatly from those predicted by classicalsteady state methods. It is here that transient two phase flow simulators offer considerablepromise if interface tracking methods can be improved. However, one must be cautious aboutthe application to real systems, It can be envisaged that in a long hilly terrain pipeline a sectionmay exist close to the slug/stratified boundary. A small perturbation in the pressure or flowratemay be all that is required to trigger a slug and change the characteristics of the whole pipeline.

In some cases it has been observed that the exit of a slug in a pipeline apparently triggers theformation of another slug. Slugs occur regularly with one, two, or three in the line at one time.This is due in part to the time taken for the liquid level to re-establish, and partly due to the de-packing effect as the slug exits, removing the high pressure drop over the slug body. Thiscauses an increase in the gas velocity which can trigger another slug. This is illustrated inFigure 7.11 which shows the passage of slugs throughout a 400 m long, 8” diameter, horizontalair/water test facility. The Y-axis shows the identification number for 30 Light Emitting Diode(LED) slug detector probes, which are located along the top of the pipeline, and give a binary‘on’ output when water wet and a ‘off’ signal when in air. As the number of slugs in the lineincreases the influence of each slug exiting diminishes and slug formation becomes less orderly.The topography of the pipeline can exaggerate this effect if uphill sections are present at theoutlet. This can increase the pressure fluctuations due to hydrostatic effects, and can lead tosevere flowrate fluctuations such as demonstrated by severe slugging. The random type ofslugging with considerable decay and coalescence is demonstrated in Figure 7.12.


Section 7. Transient Flow BP Multiphase LIesi@ Manual

v., = 0.33 m/s v... = 2.6 mls

1 III I m

2e nl3 I I r4u5n n -1161 07 . 111 1

6. m-l

9- r - l n10 ~11 12 I r13 14 r-l m r - l

L.e.d. probe :z -number m r-7

17 I16. II m

19 _20 m---21 I 1 m

2223 I

25 n26 fl27. ll 1 r - l IT-

29~ 1 m-l N

3 0 rn nn n n n n n I nllnn 0 r

I I I I I I I I I0 50 100 150 200 250 300 350 400 450 500

Time (Seconds)

Figure 7.11 Periodic single slug generation

V,, = 0.66 m/s V, = 2.6 m/s

1234

56769

10 n n r l n n n m n Fil nl n n

11 n n n n n n rru I n n r-7 n n

12 r - l n n nl i

13 2r-l m n-l

14 n rl n n n n r - l r - l n rlr

Led. p r o b e ;; En ll n n

numbern n

17 n rl nn n r-7 nn n I n r-l nrL

16 ~n~n~nln~rllnlm~nlY n n rlr

19 n n n r20. n n l-l nl n n m n n II n21 n n n n nn u n r-l n n m n2223 n n r-l n n n n n n2 4 n n n 1 r l n rl n r - 7 r25 n n n n n n n r - l n-l26 , n n r-7 n n r-l n n rl2 7 -I n m n n n n n n ,26 I I29 7rl-n r - l n n n n3OJIU 1 r-l I n r-l n r-7

0 2 0 4 0 6 0 60 100 120 140 160

Time (Seconds)

Figure 7.12 Erratic high frequency slug generation


In some cases the bubble length is greater than the pipeline length. Here it can take some timefor the liquid level to be re-established before the next slug is generated. Slugs generated inthis way can often have the longest lengths, as the equilibrium liquid content in the pipeline canbe high. The slug grows as it sweeps up the liquid in front of it, leaving behind a thin film whichmay take a considerable time to re-establish. Such phenomena may be well predicted byPlAC when the slug tracking model has been implemented. In the meantime it is intended toimplement a simple maximum slug length model in MULTIFLO. This has been shown to givegood agreement with the slugs generated in the Endicott pipeline under these conditions.

PLAC has been shown to be capable of accurately predicting severe slugging, and can simu-late slug generation caused by slug exits (see Section 7.4.2). However, the models in PLACrequire improvements in order to properly track the slugs and to simulate slug growth. Thecode will in general give reliable results when the flow is dominated by hydrostatic effects, sueas severe slugging and flows in flexible risers. More work is required in order to predict slugflows in horizontal lines. Here the existing design methods should be used.

:h

Some initial studies using PLAC to predict slug distributions have shown promise. For exampleFigure 7.13 shows PlAC predictions of the slug length distributions for the Sunbury tests fromthe 6” air/water horizontal pipeline, the corresponding measured slug length distribution isshown for comparison in Figure 7.14 and gives a mean slug length of 4.4 m and a maximumlength of 12 m. It is seen that the agreement between the simulation and the experiments isgood. Under these conditions of high frequency slug flow PtAC predicts tall waves rather thanslugs. A tracking method is being developed to increase the velocity of these waves to the slugvelocity in order to model the slug growth from liquid sweep-up in front. Work is continuing inthis area.

80

60

Frequency 4.(no.)

30

20

10

0

Y

Measured mean lengh = 4.4mMeasured maximum length = 12.0m

I8.57 I

8.88

Length (m)

Figure 7.13 PLAC comparison with RCS slug distribution


Section 7. Transient Flow BP Multzphase Design Mmwul

60

54

48

42

38Number 3.of slugs

24

18

12

6

Measured mean length = 4.4mMeasured maximum length = 12.0m

Transformed Data

Figure 7.14 Experimental slug length distribution

The extrapolation of existing design methods to predict slug flow in hilly terrain pipelines may berisky, and it is here that dynamic simulation may be of use. This is an area where validationwork is underway. PLAC has been compared with other Sunbury experiments performed with apipeline dip configuration on the 6” air/water rig, the rig set-up is shown in Figure 7.15. In thesetests a known quantity of water was poured into the dip and allowed to settle. The air flow wasthen turned on at a constant rate and the flow behaviour observed. The tests were repeatedfor a range of initial liquid quantities and air flowrates. At low gas flowrates the slugs are gener-ated at the dip where the liquid depth is the greatest and leads to a higher local gas velocitywhich is sufficient to produce waves that grow into slugs. This is illustrated in Figure 7.16. Athigher gas velocities the slug generation point moves downstream of the dip and the slug gen-eration is more random. This is seen from the densitometer traces illustrated in Figure 7.17.The flow pattern map generated by the matrix of liquid volumes and gas velocities is shown inFigure 7.18, where the solid line shows the limiting gas velocity at which liquid is removed fromthe dip. At this point the liquid is drained and measured and gives an indication of the equilibri-um liquid holdup. This is compared with the terrain induced slug flow model described inSection 4, where the agreement is seen to be generally good.

Air 6 densitometert 1.3”

4 4 -4 w20m 46m 66m

Rig topography

Figure 7.15 Experimental set-up


Oil

Oil

Figure 7.16 Dip slug generation

Test 8 (509), 1.03 m/s, regular generation frequency

:+J-ud00 11 II II II II II

00 12 24 36 48 60

Test 9 (510), 1.97 m/s, less regular

Test 10 (51 l), 3.04 m/s, even less regular

-8 -

0000 12 24 36 40 60

0 0 ;; II II II II II00 12 24 36 40 60

Figure 7.17 Densitometer traces for various gas velocities



600 -

500 - n =

x Liquid Losso Random Slug+ RippleA Static

400 - n

Liquidv o l u m e 300- ’

litres

n Slug

0 Wavy-.-. Terrain induced

slug model

I I

0 1 2 3 4 5 6 7 8 9 10 11 12

Air Velocity m/s

Figure 7.18 Pipeline dip experiments-flow pattern map and equilibrium holdup

Initial comparisons with PLAC have been encouraging as illustrated in Figure 7.19, which showsthat PLAC generally predicts the gas velocity at which liquid is removed from the system andthe equilibrium holdup, however PLAC tends to underestimate the holdup at the higher gasvelocities. Figure 7.20 shows that the predicted slug frequency is generally lower than that mea-sured. However, the results are still in reasonable agreement. This is because PLAC simulateswaves rather than slugs, Figure 7.21 shows the oscillations produced by PLAC for case of 317litres of water with an air velocity of 0.5 m/s.

600

500

400

Liquidvolume 300

litres

200

100

0

x Experimental0 PLAC

I IIll I I I I I I I1 2 3 4 5 6 7 6 9 10 11 12

Air Velocity m/s

Figure 7.19 Comparison between PLAC and measured liquid removal limit


-

800

700c’ 6008g 5 0 0

0 II I I I I I I I I I I0 1 2 3 4 5 6

Gas velocity (m/s)

7 8 9 10

Figure 7.20 Comparison between PIAC and measured slug frequency

Figure 7.21 PUG holdup oscillations for 317 I of water and 0.5 m/s air velocity

TRANFLO can also be used to simulate slugging in dips, and was successfully employed toinvestigate the water motion in the dips of the Amethyst gas pipeline following a failure of thepig launcher. In this case MULTIFLO was used to predict flowing in-situ gas velocity andTFIANFLO was used to assess whether there was sufficient motion to provide adequate corro-



sion inhibitor transport to the dip. Figure 7.22 shows three-phase TRANFLO predictions of theslugging in a dip with gas, water, and condensate present simultaneously, the contours repre-sent the water holdup fraction. The oscillatory motion in the dip (distance = 1 OOm) was thoughtto be sufficient to effectively distribute the corrosion inhibitor.

Time(s)

80.0 -

60.0 -

olo 50.0 100.0 150.0

Distance (m)

200.01

250.0

Figure 7.22 TFIANFLO simulation of three phase flow in a dip.

In practice oil and gas pipelines will usually reach or attain equilibrium first at the start of thepipeline, hence slugs can be initiated near the pipeline inlet. It is possible for inlet flowrate fluc-tuations to initiate slug formation. Such inlet flowrate oscillations may result from slugging wells,gas-lift operations, and test well changes for example, and should be suspected as possiblecauses for flowrate perturbations which generate slugs in otherwise stratified flowing conditions.When using transient flow simulators it is worthwhile to consider the effects of practical flowrateperturbations on the result. This is facilitated in the transient code TFiANFLO which allows theuser to enter noise parameters at the inlet. The amplitude and the frequency of the inlet pertur-bations have been shown to have a large influence on the simulation, and have been useful to‘tune’ the code. Some good results have been obtained using this technique. However, it isnot possible at present to determine what inlet perturbation is required to give the right answerapriori.


7.4.2 Effects of topography on process plant

For gas/condensate systems the point of slug formation usually occurs further down thepipeline, depending on the liquid condensation rate and topography. Slugs are not usuallyformed at the start since the pipeline conditions can be above the dew point. For typical off-shore gas condensate pipelines terminating on land, the greatest uphill gradients can be at thebeach, making this a possible location for slug formation. This phenomena was observed indynamic simulations of the Bruce/Frigg pipeline system where liquid drop-out occurred whenBruce was shutdown. This resulted in a pressure decrease along the pipeline which gave riseto an increase in the liquid drop-out rate due to retrograde condensation. Predictions usingPLAC showed that following a 12 hour shutdown it required around 48 hours to re-pack thepipeline (Figure 7.23). During the shutdown liquid drop-out occurred in the pipeline, which waspartly swept out when Bruce gas was re-introduced into the system. PL4C shows that duringthe start-up phase some of the liquid evaporates as the pressure increases. That liquid whichis swept out arrives as a train of 700 bbl slugs which are easily handled by the process plant(Figure 7.24). Another result from the PLAC simulations was the fact that the 12 hour shut-down was not long enough for equilibrium conditions to be obtained, hence the liquid content isless than the equilibrium value. This is contrary to the results obtained using steady state analy-sis of the holdup change for the cases with and without Bruce gas which indicated that around4000 bbls of liquid could be removed from the pipeline.

The user should take care with this type of analysis to make sure that the rate of condensationand evaporation is realistic, as it is possible that the simple nature of the phase change modelin PIAC may give rise to large errors. For example, the use of a single composition for the fluidmay mean that the condensate is re-evaporated at a much lower pressure than would beexpected in practice because the condensed liquid would have a heavier composition than thefeed stream.

Cell position = 0 Pipe component 12 hour shut down

X10’

Totalvapour

(kg)

Time (Seconds)

Figure 7.23 Pipeline vapour content



Cell Dosition = 0 PiDe component 12 hour shut down

X105X105

Totalliquid(kg)

8.4/ I

8.2 -

8.0 -

7.8 -

7.6 -

7.4 -

7.2 -

7.0 -

6.8 -

6.6 -

6.4 -

6.2 -r I I I, I I I I I I I1 I I I

0 2 4 6 8 10 12 14 16 WW20 22 24 26 28 30

Time (Seconds)

Figure 7.24 Total liquid content showing slugging during re-start

It is seen that there can be numerous transient effects that are caused by the topography.These can be difficult to predict using steady state analysis techniques. The lack of an inclina-tion term in the present slug sizing methods and the inability to allow for hysteresis effects areboth examples and HillyFlo goes part way to addressing these problems. Transient simulatorscan in theory account for both phenomena. However, in slug flow improved models arerequired to track the slugs. The effects of slugs discharging should be well suited to transientanalysis provided that slug generating mechanisms can be clearly defined. However, if a sys-tem has numerous slug initiation sites, and is susceptible to small perturbations, then one mayexpect that the long term accuracy of the simulations may be suspect.

Codes such as PLAC and TRANFLO have produced good results with pipe emptying problemsand slug generation with topographical changes. However the slugs produced by PLAC are notreal slugs. PLAC can generate slugs produced by gravity effects but is limited in the way thatthe slug density and velocity are modelled, hence - more work is required.

Dynamic flows produced by topographical changes are important to the design of processplants, as slug characteristics are changed while negotiating the platform riser and pipeworkupstream of the receiving vessel. If the slug is long, the additional hydrostatic head requires anincrease in the pipeline pressure to force the slug up the riser. As the slug is produced into theseparator it may accelerate due to the reducing gravitational resistance and the reduced fric-tional length (Figure 7.25). This can cause large velocity increases which can impact on theprocess plant control, and can also give rise to large loads on the vessel internals. This is dis-cussed more fully in Section 10 of the Multiphase Design Manual.


ISection 7. Transient Flow

Psep = constant

Psep

-Slug slows down as line packsto generate hydrostatic head

P+P head

Psep

Excess pressure and reducing frictional lengths lead toslug acceleration

Slugvelocity

Time

Figure 7.25 Velocity increase during slug reception

Process plant dynamics can be simulated by a number of codes. BPX Sunbury presently usesthe SPEEDUP code which has also been interfaced with a simple slug hydraulic model toenable the simulation of slug effects on process plant to be investigated. The slug size is aninput to the simulation. However, it is possible to see the effect of the slug catcher pressure onthe passage of the slug. Classical control schemes generally reduce the gas compressorspeed if the separator gas flow is reduced, hence when the slug is produced the compressor isdecelerating. A better solution can be to increase the compressor speed when the gas flowdrops off, hence reducing the separator pressure and sucking the slug up the riser. This hasthe effect of reducing the gas starvation period and also means that the compressor is acceler-ating when the gas surge occurs.

A recent study using SPEEDUP to investigate control strategies to limit the impact of slugdynamics on the process plant has shown that gas outlet flow control can reduce the peak inthe gas surge. As the slug is received the gas flowrate reduces, by opening the gas outlet con-trol valve the pressure in the slug catcher is reduced, hence sucking the slug in and allowingsome scope for increase in the slugcatcher pressure to absorb some of the gas flowrate surge.

PLAC can also model a simple separator with controls, including valves. However, it is notcapable of simulating the compressor in detail. In the future is is hoped that the transientpipeline and process dynamic simulators will be coupled to facilitate integrated system analysis.A simple interface already exists between OLGA and D-SPICE.




LEFT BLANK



7.4.3 Hilly terrain pipeline shutdown and restart

Experience with the operation of the Cusiana Phase 1 system in Colombia has shown thatmaximum flowline pressures may be generated by flowline start-up. The longest multiphaseflowline in the Phase 1 gathering system is the 20” diameter, 20 km long flowline that transportsthe production from wellpads T and Q. The mountainous topography from the remote well T tothe Cusiana Central Processing Facilities is illustrated in Figure 7.26. When the production fromT was stopped the shut-in pressure at T pad was around 1300 psia compared to the arrivalpressure of 640 psia. The high back pressure resulted from segregated liquids draining and fill-ing the pipeline dips, however the liquid levels were suspected to be different, hence causing anon-equaiised hydrostatic equilibrium condition. A simple 2” air/water model of three pipelinedips was constructed at Sunbury to demonstrate what happens during the shutdown andrestart and it was confirmed that it is possible for the liquid to drain into the dips with unequallevels, hence producing a back pressure due to the hydrostatic difference in the levels. Figure7.27 shows the topography of the test rig as modelled by PLAC whereas Figures 7.28 and 7.29show the PLAC simulations of the liquid draining and inlet pressure variation.

ElevationElevation

(4(4

TtoCPFTtoCPF PipePipe

460460

420420

380380

360360

320320

30013001 , , , , , , , , , , , , , , l&l& , ,

00 2000 4000 6000 8000 10000 12000 14000 16000 16000 200002000 4000 6000 8000 10000 12000 14000 16000 16000 20000

Distance (m)Distance (m)Pipe topographyPipe topography

Figure 7.26 Cusiana T pad to CPF topography used in PLAC

Issue2 Issue2 -- December1994 December1994 37


Elevation(ml

Hilly terrain rig-air and water Pipe

titi

6

0.2 -

0.0 I I I I I I0 5 10 15 20 25 30

Distance (m)Pipe topography

Figure 7.27 Topography of of hilly terrain test rig

Liquid holdup(vol fraction) I

,,A \““i,I

0102 '-- Time (sets)

Figure 7.28 PLAC simulation of liquid draining into dips

-



Hilly terrain rig-shutdown Pipe

105105

PressurePa)

102 104 106 106 110 112 114 116

Time (Seconds)Cell number

Figure 7.29 PLAC simulation of back pressure due to liquid settling in dips

While it is seen that PLAC can simulate the liquid settling behaviour, the comparison with theexperimental results is not always good. This is because PLAC does not model slugs properlyand hence the holdup in the uphill sections is sometimes underestimated, hence the initial con-dition is not always correct.

The highest back pressures can be generated when the flowline is restarted as it is possible forthe liquid in the dips to be displaced downstream before the gas is able to penetrate the dip,and this increases the difference in the hydrostatic levels. Once the gas penetrates the pres-sure is reduced as the gas lifts the liquid out of the dip. One can imagine that the start-up pres-sure can be large if all the liquid in the dips is displaced simultaneously.

A simple method has been established to estimate the pressure rise during the start-up of hillyterrain flowlines. For long hilly flowlines the total uphill elevation changes may be very large andhence the total theoretical hydrostatic head may lead to very conservative start-up pressures. Itis hence recommended that the flowlines are started up by unloading the sections closest tothe facilities first. The pressure can be estimated by summing the uphill elevation changes andcalculating the hydrostatic head for the section being started-up. In practice experience withthe start-up of Cusiana T pad indicates that the actual hydrostatic pressure generated wasaround 213 of the maximum theoretical value, however this is very dependant on the particularsystem.

If the analysis of start-up pressures is required to check the design pressure of the flowline,then the topography should be investigated to determine if there are low points where the pres-sure can be higher than at the well pad. This is the case with the T-Q flowline where the maxi-mum pressures are generated in the valley just downstream of the wellpad.




LEFT BLANK


7.4.4 Pipeline-riser interactions

A special topographical effect is the interaction between a pipeline and a vertical platform riser.This can give rise to a phenomenon called severe slugging if the pipeline inclination is downhillat the base of the riser and if flowrates are low enough. Severe slugging has been discussed inSection 3.3.3 and will not be described again here. However, it is useful to demonstrate thatdynamic simulation can be employed to accurately predict the size of severe slugs and thecycle time, enabling topside process plant to be sized. One of the validation cases for thePLAC code was a comparison with severe slugging experiments carried-out by BP at Sunbury.The test rig consisted of a 45m flowline inclined downhill at 2” followed by a 15m vertical riser.The rig operated at near atmospheric conditions and the test fluids were air and water. Aschematic of the rig is shown in Figure 7.30 where the pipeline inner diameter is 2”. Figures7.31 and 7.32 show the measured slug size and frequency for a range of fluid superficial veloci-ties.

To separatorI

1 15m riser

Gas in

Air

Water

50m pipeline

2” Mixing

PIPELINE RISER TEST RIG *l 2 in diameter

Figure 7.30 Schematic of severe slug test rig.

334

434 232 276,8598

296 167 168167

215148 74

ri 0 . 0 1 /,

0.1 1 10

Superficial gas velocity (m/s)

Figures 7.31 Experimental slug volume (% riser)



5757

155 78155 787676

87 87 3’157 97 97 8’ 8484

226226127 76127 76

231231

Superficial gas velocity (m/s)

Figures 7.32 Experimental cycle time (s)

Cell position

.Flowline

0lime(units)

Figure 7.33 PLAC holdup profile prediction.

Figure 7.33 shows the liquid holdup profiles predicted by PIAC for gas and liquid superficialvelocities of 0.44 m/s and 0.43 m/s respectively. The Figure shows the liquid build-up andblowdown. Figure 7.34 illustrates the intermittent nature of the outlet liquid mass flowrate,showing the rapid increase in the flowrate as the tail of the slug accelerates up the riser. Forthis case the experiment gives a severe slug size of 434% of the riser volume and a severe slugcycle time of 155 seconds. The PLAC simulation gives 430% and 150 seconds respectively. Itshould be noted that it is important to correctly model the volume of the system in order tomatch severe slugging cycle times. In the case of the Sunbury experiments the model in PlACincluded a section of air filled pipe at the inlet which was used to model the line supplying air tothe mixing section.


55050 150150 250250 3.503.50

TmeTme (s)(s)

LL+450450 550550 650650

Figure 7.34 PIAC prediction of outlet liquid mass flowrate.

A similar type of severe slugging can also be observed when flexible risers are employed, par-ticularly if the lazy ‘s’ configuration is adopted. The analysis of the flows in these curved config-urations is complicated by the possible bi-directional motion of the flow, and the interaction withthe complex geometry. Here transient simulators show promise in being able to predict theflows, whereas the scale-up of laboratory simulations to practical situations is uncertain.Example 7.4 shows a comparison between PLAC and some laboratory experiments of severeslugging in a catenary riser.

The relationship below can be used to estimate the critical liquid velocity at which severe slug-ging occurs in a pipeline-riser system. If the liquid superficial velocity is above this value severeslugging is unlikely to occur.

V,, = Vsg Psep / Lb, - PJ g (1 - H,) Sin 01

where:

P=P=P

= pressure at flowline outlet (N/m*)L = line length upstream of riser (m)p, = liquid density (kg/m3)p, = gas density (kg/m?H, = average liquid holdup in lineg = acceleration due to gravity (9.81 m/s*)I3 = inclination of riser from horizontal




LEFT BLANK


7.4.5 Transients Caused by Looped Pipeline Systems

Pipeline looping is a common practice to increase the capacity of systems transporting singlephase oil or gas. The parallel loop may be installed along the whole or only part of the pipelineto increase the effective flow area, and hence to remove a bottleneck in the system or to facili-tate an increase in the flowrate. In single phase flow the analysis of the hydraulic performanceof the loop is usually determined using a single pipe with an equivalent diameter. This methodis also often applied to the looping of two-phase pipelines using simple tee junctions, whereonly one leg with half the flowrate would be analysed if both pipes in the loop have equal diam-eters, for example. Inherent in such an analysis is the assumption that the gas and liquid splitsin equal proportions in both of the legs. However, in practice this is not usually the case.

The two-phase flow split at pipeline tee junctions is complex, and at present there is no univer-sal method for determining the flows to each leg, due to the dependence on the geometry ofthe tee, and the various gas and liquid flowrates and physical properties. However, it is appar-ent that in most cases the flow split at a side arm tee junction is usually very different from beingequal. See Section 11 of this manual.

At very low liquid fractions in stratified flow typical of gas transmission systems with a smallamount of liquid dropout or carryover, it is possible for the liquid in the stream to follow the gasinto the side arm of the junction, and hence the run remains virtually dry. Instances have beenrecorded where liquid has consistently taken one path in a complex gas transmission network,resulting in overfilling pipeline drips, and causing an interruption in the supply of gas, althoughthe overall liquid content of the liquid transmission system is very low. At the other extreme, inslug flow for example, the inertia of the liquid causes the majority of it to flow forward into therun, whereas all the gas may flow through the side arm. In fact some simple tee junctions areused as in-line separators.

It is hence seen that the flow split at a simple tee may be very different from equal, and hencethe flowrates in the individual legs will be unequal. However, since in a loop the lines rejoindownstream, the number of possible solutions is reduced to those flow splits that give rise tothe same overall pressure drop in both pipelines.

Transient flows can result from the operation of looped pipelines since it is possible for a num-ber of solutions to exist for the flow split. For example, high flow in one leg causing a high fric-tional pressure drop may be balanced by low flow in the other with a high hydrostatic compo-nent and high holdup. Small inlet flowrate perturbations at the tee, or changes in operatingconditions, may cause the flowing conditions in each leg to change, with the possible removalof excess liquid in the form of a slug. In the extreme it is possible in pipelines with a large uphillelevation for a manometer effect to exist. Here the flowing pressure in one leg is balanced by astatic column in the other. A change in operating conditions can cause the static liquid to beswept-out. This type of phenomena has been known to give rise to operational difficulties withgas transmission pipelines laid over hilly terrain.

Problems with parallel loops are well documented in the power generation industry, whereinstabilities can occur in a bank of boiler tubes, for example. The total flow into and out of theheaders is normally constant, as is the pressure drop. However, flows in the individual tubescan vary considerably. In such situations the analysis is complicated by the addition of heattransfer effects and steam generation, which can give rise to multiple solutions for the locationof the boiling front. So called ‘parallel channel’ instability in the boiler tubes can cause corro-sion problems or dry-out leading to premature failure. Experience with such situations hasshown that adding flow restrictors to the inlet header has a stabilising effect, whereas throttlingthe outlet can exacerbate the problem.



The method proposed to analyse possible transient problems caused by operating loopedpipelines is along the lines outlined by Gregory and Fogarasi (Ref 7.1). This involves calculatingthe two-phase pressure drop in each leg for a range of liquid and gas flow splits. A graphicalsolution is employed to determine the conditions where the pressure drops in each leg areequal, and hence the possible operating regimes of the loop. Allowance should then be madefor flow excursions between the parallel lines and the possible displacement of liquid.

As an example consider the schematic shown in Figure 7.35

Q.oln

Qgin

Qoa GORaa Qga

cRS

GOR

Qob Go&

b Qgb

Figure 7.35 Schematic of a pipeline loop

Let OL equal the fraction of the inlet liquid flowrate in line A and 0 equal the fraction of the freeinlet gas in line A, Hence:

Qoa = aQoin and Qga = OQgin

The free gas at the inlet to the loop is given by the difference between the producing gas-oilratio and the solution gas-oil ratio at the inlet conditions.

hence:

Qga = B(GOR - RSin) Qoin

These expressions can be used to determine the effective producing gas-oil ratio to use in thetwo-phase pressure drop analysis as follows:

GORa = [Qoa Win + O(GOR - RSin)Qoin] / Qoa

hence:

GORa = RSin + O/a (GOR - RSin)


i


Similarly for line B:

GORb = RSin + (1 -I3 / 1 -a)(GOF? - RSin)

Steady state pressure drop calculations can be performed for a range of gas and liquid splitsusing the effective GOR calculated above. The value of the solution gas-oil ratio at the inlet con-ditions can be determined using the PVlTAB facility in MULTIFLO based on the inlet tempera-ture and pressure.

Figure 7.36 shows the locus of the possible solutions for a equal pipeline loop and indicatesthat the 50/50 solution (CX = 0.5) gives rise to the maximum downstream pressure, and henceminimum pressure drop, however this is not always the case. For other pressure drops thereare two flow split solutions and hence it could be possible for the flows to oscillate between thetwo solutions.

5 6 0

5 4 0

5 2 03B

z!?ii 5 0 08hE8z 4 8 0

i!3::

$ 4 6 05508

4 4 0

4 2 0

4 0 0

P= P=0.9 \0.7.‘. -1-- LINEA

, 0.1

l * -1-1 LINE Bj/ 0.3

D.5 * ‘,‘..**.**.**.**

0 . 3 .**l **

l .l * l .

I I I I 10 0 . 2 0 . 4 0 . 6

(Y

Figure 7.36 Locus of solutions for equal pipeline sizes

0 . 8 1 .o


ISection 7. Transient Flow BP Mult@hase Design Manuul

600

500

400

300

200

- Line A

---- Line B

0.1

I I I I I I

0 0.2 0.4 0.6 0.8 1.0

aFigure 7.37 Locus of solutions for unequal pipelines

Gregory et al (Reference 7.1) suggests that the existence of a manometer leg effect can bechecked if the net elevation increase over the loop is such that the maximum possible hydrosta-tic head that could result from the liquid phase static in the pipeline is greater than the pressuredrop that will result from the total throughput of gas and liquid flowing in one of the parallellines. This may not be wholly true as a symmetrical hill would have a net elevation change ofzero. However the potential exists for the uphill section to contain liquid and provide a manome-ter effect. It is hence recommended that the manometer effect check be carried out using thesum of the uphill elevation changes. If one of the legs contains static liquid it is obvious that theloop has in fact not increased the capacity of the system as the flow only occurs in one leg.

In the above example the pipes in the loop are identical and hence the locus of possible solu-tions is symmetrical. This is not the case when the pipes are different as illustrated in Figure7.37 for a IO” and 12” loop from Gregory’s paper. Here the locus of the possible solutions isnot symmetrical and the minimum pressure drop occurs with a liquid split of 65/35.

Example 7.5 shows the analysis technique applied to a proposed multiphase pipeline loop overmountainous terrain in Colombia.

The main points to remember about the hydraulics of looping two-phase pipelines are as fol-lows:


1. If the pipes in the loop are identical, the assumption of equal split may lead to anunder prediction or over prediction of the pressure drop in practice, as equal splits areunlikely. For some conditions equal split can give a minima in the locus and for othersa maxima.

2. A range of flowing solutions are often possible which give equal pressure drops.While this may give rise to steady outlet flowrates, the natural perturbation producedby multiphase flows, and the complex characteristics of flow splitting, can lead tooscillations between the legs, and consequent transient liquid sweep-out.

3. Although the outlet flowrate from the loop may be steady at one condition, a smallchange in the operating conditions can lead to transient liquid sweep-out as the newoperating point stabilises.

4. A check should be made for manometer effects based on the sum of the uphillelevation changes, although this may not give the maximum holdup change as a pipefull flowing case may also exist. If manometer legs occur the loop will not increase theflowing capacity.

5. Static liquid legs may have implications for corrosion.




LEFT BLANK

50 Issue 2 - December 1994--~~

Transients due to pigging pipelines

Introduction

A method for estimating the slug sizes resulting from pigging pipelines is presented in section 4.This approach assumes that the pig travels at the in-situ mixture velocity, which is not alwaysthe case in practice. The motion of the pig may vary as a result of the compressible nature ofthe fluid in the pipeline, which allows pressure and velocity excursions to arise from the varia-tions in the pipeline inclination and the friction between the pig/pig train and the pipeline. Forexample, the pig may require additional driving pressure in order to push the slug over a hill orto compensate for an increasing frictional resistance as the slug grows in length. If the system is‘spongy’ the slug velocity may decrease or even stop as the fluid packs behind the pig to pro-vide the additional driving pressure. In downward sections the head recovery can cause the pigto accelerate and transfer the excess pressure into friction loss.

If the pig stops, then it is possible to observe a ‘stick-slip’ phenomena caused by static anddynamic friction effects, where more pressure is required to get the pig moving than is requiredto keep it moving. The additional pressure generated after the restart can cause the pig toaccelerate, de-packing the line, and hence stopping again.

These phenomena can have a large effect on the time required to pig a pipeline, on the dynam-ic pressures generated, and on the fluids produced. This may affect the design and operabilityof the system. In addition, it is sometimes important to predict how long it takes to re-establishthe liquid holdup behind the pig, and hence to re-distribute corrosion inhibitors.

Simple pigging transients can be investigated using the OLGA and PLAC transient two-phaseflow codes, both of which allow for a variable leakage past the pig during the pigging process.The PLAC model is a recent addition and has yet to be properly validated, but is used here todemonstrate the transient effects caused by pigging pipelines.



7.5.2 Transient pigging example-pigging a hilly terrain pipeline

Figure 7.38 shows the topography of a hypothetical hilly terrain pipeline which is 200 mm indiameter, 10 km long, and includes a 50 m high hill followed by a 50 m deep dip. A high GORcomposition was used and the simulation run to a steady state with an inlet mass flow of 2 kg/sand a delivery pressure of 10 bara. A pig was launched into the pipeline at time 20100 s and itwas estimated that the pig transit time would be around 2900 s based on the average mixturevelocity in the pipe. Figure 7.39 shows the pig position with time, and indicates that the pigmotion was not steady and displayed three regions of steady motion and two regions of slowermotion, and in some cases reversal. The actual pig transit time is 5900 s, which is almost twiceas long as estimated from the average mixture velocity. This discrepancy results from the factthat the pig slows down in the uphill section as the pressure builds-up.

50

25

-25

-50

2000 4000 6000 8000 10000

Distance (m)

Figure 7.38 Topography of hilly terrain pipeline

Figure 7.40 shows the holdup profile throughout the pipeline for various times and indicatesthat before the pig is launched the holdup in the upward inclined sections is around 35 %,whereas in the horizontal and downward sloping sections it is around 2 %. The pig slug isshown just after the pig has passed the top of the first hill and indicates that the holdup isaround 95 %, and that the slug is less than 500 m in length. The holdup profile at 40,000 s indi-cates that the liquid content of the pipeline is still building-up 4 hours after the pig has beenreceived. Figure 7.41 shows the pressure at the pipeline inlet and indicates that a pressure riseof 1.3 bara is required for the pig to negotiate the first 50 m hill. The liquid full hydrostatic headwould be equivalent to 2.8 bara, hence the generated pressure is consistent with the observa-tion that the pig slug half fills the upward inclined section.

issue 2 - December 1994


1 0 0 0 0

8 0 0 0

-5 6 0 0 0E. -

.%2im 4 0 0 0.-a

0

2 4 0 0 0 2 6 0 0 0 2 8 0 0 0

Time (s)

Figure 7.39 Pig position with time

- - 1 0.2002E 0 5

- 0.2376E 0 5

- 8 1 0.3993E 0 5

0 2000 4000 6000 8000 10000

Cell distance (m)

Time (s)

Figure 7.40 Pipeline holdup profiles



2.6

2.4

1.4

1.2

I I I I I I21000 24000 28000 32000 36000 39000

Time (s)

Cell number

Figure 7.41 Pressure at pipeline inlet

Although the new pigging model in PLAC is simple in nature, it is seen that the general motionof the pig can be very different from that expected by simple steady state analysis. Future mod-ifications to the model are required to improve the treatment of the pig/pipewall friction andcode stability.


7.6 Pressure surge transients in two-phase flows

Pressure surge transients in single phase flows are well understood, with computer packagesavailable to predict the surge characteristics of complex pipeline networks with componentssuch as pumps, valves, and surge suppression devices. A computer program called FLOW-MASTER is extensively used by BP Exploration for this purpose. When a two-phase flow isintroduced into the system the analysis is made far more complicated since the speed at whichthe pressure waves travel is modified by the gas fraction. In addition, the discontinuous fluidinterfaces in two-phase flow provide additional locations for pressure waves to be reflected andattenuated. Such is the complication that no general purpose pressure surge code exists fortwo-phase systems. There are however a number of ways that two-phase surge pressures maybe analysed with FLOWMASTER and these will be outlined below. We will begin with a briefintroduction to single phase pressure surge phenomena, followed by a discussion of the effectsof two-phase flow, supported by some examples.

-




LEFT BLANK


--

7.5.1 Single phase hydraulic surge analysis

When a velocity change occurs in a flowing fluid some of the kinetic energy is converted intopressure energy giving an effect called hydraulic surge or water hammer. In most single-phasepipelines this can be caused by delivery valve closure and pump start-up or shutdown. There isnormally the potential to create pressures significantly above or below those occurring duringsteady conditions giving a potential for overpressuring and subsequent rupture or the genera-tion of low internal pressures leading to pipe collapse. In the case of a closing valve the effect ofthe increased restriction is to reduce the flowrate. At the instant of the increased restriction theflow further upstream of the valve continues at its previous rate which leads to the conversion ofkinetic energy into potential energy. The slowing or stopping of the fluid travels as a pressurewave against the previous flow direction. In simple systems the highest pressures are generat-ed when the flow is stopped instantaneously, however in practice the magnitude of the surgepressumis influenced .bL’-p:esst;re-\N~~~reflecte .-d fromthe-other end of the system. Thesewaves require a finite time to return to the location of the flow stoppage, which is called thepipeline period. If the flow has completely stopped within the pipeline period, then an ‘instanta-neous’ surge pressure will be experienced.

The pressure wave time period will be modified by the formation of vapour cavities that occurwhen the pressure falls below the vapour pressure. The pressure waves also experience asudden change in magnitude when passing a change in the pipeline diameter. A pressuresurge can also induce a flow in another line connected to the system, which can lead to amplifi-cation of the original surge if the effect of stoppage of this induced flow is added to the originalpressure surge.

The interaction of flows and pressures with pumps, line friction and other modifying factorsmake it impractical to calculate anything but the most simple systems without computer pro-grams such as FLOWMASTER. However, hand calculations can be useful to check if a systemhas the potential to produce surge pressures in excess of those the line can withstand usingthe equation below:

P = -paAv

where:

P = pressure riseAV = reduction in flow velocitya = speed of pressure wavep = fluid density

If the fluid is stopped instantaneously the maximum head or so called ‘Jacowski’ pressure riseresults. In a practical system the maximum head may be generated if the fluid is brought to restwithin the pipeline period. For example, consider a pipeline discharging through an outlet valvewhich closes fully. The full Jacowski pressure rise occurs if the valve closes within the timetaken for the pressure wave to travel to the other end of the pipeline and back. This is thepipeline period and is given by:

7 = 2L/a

-


where:


7 = periodL = length of the pipeline

This method may be used to check the potential for overpressuring the pipeline, but may notgive the maximum pressure if networks are involved due to amplification at the dead ends.

There are several operating changes that can cause flow changes and hence pressure tran-sients. The most common are valve opening/closing and pump start/stop.

In the valve closure example the shape of the pressure/time variation will depend on theflow/time change which caused it and thus on the characteristics of the valve. Typical flowcharacteristics of ball, globe, and gate valves are shown in Figure 7.42. It is important to notethat the last 10% of the valve closure has the most significant effect.

Relativelog flow

rate

Ball

G a t e

Globe

I 1 I I I I I I I I 1

0 lo 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0

Percentage valve open

Figure 7.42 Relative flowrate against percentage valve opening for typical gate, ball, andglobe valves

If the pipeline period is greater than the valve movement time then the type of valve characteris-tic will not influence the total pressure rise, although the rate of pressure rise will be important toany control devices and the mechanical load on supports. If the period is less than the valvemovement time then the shape of the characteristic will influence the peak pressure. Specialcare should be exercised in specifying valves whose effective closure time is only a fraction ofthe total movement, eg. gate valves.

Pressure falls can be as damaging as pressure rises. Often the negative head changes lead totheoretical pressures less than the fluid vapour pressure. In such cases a vapour cavity forms.Large diameter lines are more susceptible to damage by reduced pressure conditions. Morecommonly vapour cavities create damage by their collapse. This occurs if there is a situationwhich can re-pressure the line. This could be caused by refilling from an elevated section of lineor tanks or by re-starting a pump to continue operation. Often this refilling occurs at a high ratebecause the pump discharge pressure is lower than normal. When the cavity volume reaches


zero the effect is of a truly instantaneous valve closure. These positive pressures can be higherthan with normal valve closure and have step pressure changes. Therefore, they should beavoided if at all possible.

Pump start/stop has the capability of producing similar positive and negative pressure waveswith the attendant damage potential.

From the above, it becomes apparent that the significant factors in creating high surge pres-sures are the wave speed and the rate of change of flow. The wave speed is fixed by the fluidsused, although it is sometimes possible to modify the rate of change of velocity in a transient.This is particularly true of the effective valve closure time. Even valves with the most suitableclosure characteristics create most of the flow change over the last quarter of their movement.Thus, two part closures, where the first part is fast and the last part slow can be beneficial.

Where surge is a problem, it is particularly important to avoid smaller bore area ends. When thesurge pressure reaches a pipe junction it propogates along both lines. The pressure reduces inratio to the new area divided by the old area and induces a flow by the pressure difference. Atthe instant of the wave arriving at the dead end the forward velocity is stopped creating anothersurge effect on top of the pressure wave. This can lead to a potential doubling of the pressure.

One solution to reduce surge pressures is to introduce surge accumulators into the system.This is particularly effective with pump generated surges. These accumulators are vesselswhich contain a gas pocket connected as closely as possible upstream of the surge generationdevice so that when the surge occurs, the excess fluid flows into the vessel compressing thegas. By removing excess fluid (which is nearly incompressible) the line pressure is reduced. Thevolumes and costs of such suppression equipment can be quite small when the transients arerapid.

7.5.2. Wave speed in single phase flow

Obviously the calculation of the speed of pressure waves in the fluid medium is crucial to thedetermination of the pressure surge. For a single phase system the pressure wave velocity isgiven by:

ci = (K/p)“*

where:

K = bulk modulus of elasticityp = density of the fluid

Hence for water at 15 “C, K = 2.15 GN/m2 and p = 1000 kg/m3 the speed of pressure waves is1466 m/s. For an ideal gas the bulk modulus is the ratio of the change in the pressure to thefractional change in volume and hence depends on whether the compression is adiabatic orisothermal. For an adiabatic process PV = y constant and hence K = yP. Since P/p = RT theexpression for adiabatic compression of an ideal gas is:

a = (K/p)“* = (yRT)“*

Issue 2 - December 1994 5 9.~


For air at 15 “C, y = 1.4, R = 287 J/kg/K hence the wave speed is 340 m/s. The equivalent bulkmodulus of elasticity is 141.7 KN/m2 and the density is 1.226 kg/m2.

The pressure wave velocity is modified when boundaries are present such as pipe walls, freesurfaces, gas bubbles and solid particles. The elasticity of the walls of the pipe through whichthe fluid is travelling has the effect of reducing the pressure wave speed by a factor dependentupon the size, cross-section shape, and pipe material. In addition the method of pipe restraintmay also affect the result. The general equation for the wave speed in a fluid contained within athin wailed pipe of circular cross-section is:

a = 1 / [p (l/K + DWtQ]“*

where:

D = internal diametert = wall thickness@ = pipe restraint factor

For materials with a high elastic modulus such as steel or concrete, or for pipelines with fre-quent expansion joints, Q can be taken as unity. Hence for water at 15°C in a steel pipeline(E=207 GN/m2) of 200 mm diameter and 15 mm wall thickness, the wave speed is reducedfrom 1466 m/s for rigid pipewalls to 1374 m/s.

-


7.5.3. Wavespeedintwo-phase flow

The wave speed in two-phase flow can be determined by simply replacing the single phasebulk modulus and density with the two-phase equivalents, ie:

a = W&+J ‘I2 for rigid pipewalls

and:

a = 1 / [ptp (l/b, + D@tE)]“*

where:

1 (I-a) cx-=-------+

Y, KL Kg

and:

P@ = (1 - 4 PL + ‘YP,

a = in-situ gas fraction

Using air and water as an example it is seen that the wave speed is rapidly reduced by theintroduction of only small quantities of air. For example, a void fraction of only 1% is required toreduce the wave speed from 1446 m/s to 119 m/s. The table below shows the wave speed forvarious gas fractions, also shown is the relative Jacowski head rise possible compared with theliquid only case. It is hence seen that in two-phase flow the wave speed and potential surge ismuch reduced by the combined high compressibility and high density of the two-phase mix-ture.

Void fraction

t-1

0.000 2.19e+9 1000.0 1466.0 100.00.005 2.80e+7 995.0 168.0 11.40.01 1.41 e+7 990.0 119.0 8.00.1 1.42e+6 900.1 37.7 2.40.2 7.08e+5 800.2 29.7 1.620.4 3.54e+5 600.2 24.3 1 .oo0.5 2.83e+5 500.6 23.8 0.810.6 2.36e+5 400.7 24.3 0.660.8 1.77e+5 201 .o 29.7 0.410.9 1.57e+5 101.0 39.4 0.270.99 1.43e+5 11.2 113.0 0.070.995 1.42e+5 6.2 151.3 0.061 .o 1.42e+5 1.2 340.0 0.03

Two-phase Two-phasebulk modulus density

Wm2) Wm2)

Wavespeed

b-w

The wave speed is plotted as a function of the void fraction in Figure 7.43.


% potentialhead rise

t-1

61

Section 7. Transient Flow BP Multipbase Desip Mm~ual

1600

800

600

400

200

0 0.2 0.4 0.6 0.8 1.0 1.2

Void fraction

Figure 7.43 Pressure wave speed in two-phase flow


7.5.4. Two-phase pressure surge analysis using FL.OWMA!%ER

Although FLOWMASTER is essentially a single phase hydraulic surge computer program, thereare a number of features that may be used to investigate two-phase surge effects. Inherent inthe code is the ability to form vapour pockets if the pressure falls below the vapour pressurespecified for the fluid. If the pressure subsequently rises the collapse of the cavity can also bemodelled. FLOWMASTER simulates a practical system as a combination of modules that areconnected together to form a network. The modules or components represent pipes, valves,pumps etc and provide the equations for specific applications (for example valve loss coefficientas a function of opening). There are two components available in FLOWMASTER that are usefulfor two-phase flow surge calculations. These are the ‘bubbly pipe’ and ‘primer’ components.

Bubbly pipe component

The bubbly pipe component allows the user to model the wave speed in an elastic pipe with asmall fraction of free gas present (the order of 1% void fraction) which results in a variablewavespeed throughout the pipe. The gauge vapour head and the wave speed in the pure liquidare specified. The rate at which gas is released from solution is also controlled by a gas releasefactor which is a function of the per unit difference between dissolved and equilibrium gasmasses. This component is hence restricted to homogeneous bubbly type flows with small gasfractions, but may be useful to determine loads experienced as safety valves are closed on avolatile oil pipeline during a rupture, for example.

Primer component

The primer component is used to analyse the transient behaviour associated with the priming ofa part-empty pipe, and was initially developed to study the impact loads and pressures createdwhen a dry fire sprinkler system is initially primed with water. These loads are created when theliquid impacts on the sprinkler head restriction, and is similar to a slug impact. For two-phaseslug flow analysis steady state or transient analysis should be carried out using other softwareto determine the length and velocity of slugs at the actual pipeline conditions, and the primercomponent in FLOWMASTER used separately to determine slug impact loads.

The primer is a length of pipe with a restriction at each end, and the wave speed in the liquid isspecified as the initial fractional gas length in the pipe. The inlet pressure drives the liquid alongthe pipe like a piston, compressing the gas and forcing it through the outlet restriction until theliquid front reaches the end and the surge pressure is generated. By changing the speed ofsound in the liquid and the density it is possible to consider the sensitivity of the resulting pres-sure rise to the gas entrainment fraction of the slug.

Reference

7.1 Gregory, G.A and Fogarasi, M “Estimation of pressure drop in two-phase oil-gas loopedpipeline systems”, Pipeline Technology, March-April 1982, ~~75-81.




LEFT BLANK


Example 7.1 PLAC simulation of shutdown and restart ofPompano subsea wells

Pompano is a development in the Gulf of Mexico consisting of a fixed platform in 1300 ft ofwater, with a proposed remote 18 slot template located 4 miles away in a water depth of 1800ft . The template was originally going to be connected to the host platform by two 6” pipelinesallowing flexible operation as the flowrates change, and facilitating pigging from the platformusing processed oil. The first phase of the development, now completed, involved the installa-tion of the platform and start-up of the platform wells, which are expected to produce a maxi-mum of around 18 mbd. The second phase of the development will involve installation of a sub-sea template and flowlines. However it was required to make provision for this during the instal-lation of the first phase. The maximum oil production from the subsea wells was expected to be12 mbd and slug catchers were required to be sized for the platform reception facilities.

The slugs produced by pigging the pipeline were to be handled by operational procedures andwere not used for sizing the slug catcher. The predicted mean and maximum normal hydrody-namic slugs were 6 bbls and 16 bbls respectively, and were well within the feasible slug catchersize, which was in the region 50-I 00 bbls. It was hence necessary to consider the slugs pro-duced by flowline rate changes which could be quite frequent as wells are switched to test andas turndowns are accommodated. A worst case was considered to be start-up from 1 well tofull production, giving a flowrate increase from l-l 2 mbd. This was modelled with PLAC byrunning the simulation at 12 mbd for 13000s to give a steady state. The flowrate was thenramped down to 1 mbd over 60s left at this rate for 5000s then increased to 12 mbd over60s. The predicted outlet liquid flowrate profile is shown in Figure 1 and indicates that the liquidflowrate drops off when the flowrate is reduced after 13000s. The plot shows a large overshootwhen the rate is increased again at 18000s. The oscillations decay as the final equilibrium pro-duction rate of 12 mbd (equivalent to 17.7 kg/s) is approached. The peak production rate is 82kg/s which is equivalent to 51 mbd, and is hence over four times the final equilibrium produc-tion rate.

901

8 0

3 60&Y2 50

5= 4 0I

zn 3 0

s.-1

2 0

1 0

0

- 5

10000 1 2 0 0 0 14000 16000 18000 2 0 0 0 0 2 2 0 0 0

Time (s)

Figure 1. PIAC simulation of outlet liquid flowrate

The outlet gas rate is shown in Figure 2 and indicates that the main gas surge occurs after thesecond slug.



9

7

6

VAP 5mass

flow outWW 4

3

2

1n1200012000 1400014000 1600016000 1800018000 2000020000 2200022000

Time (Seconds)

Figure 2. Predicted outlet gas flowrate

The liquid holdup plot in Figure 3, shows that the initial holdup is around 40% in the inclinedflowline (cells l-l 8) and around 20% in the vertical riser (cells 19-39). During the low flow con-dition from 13000s the liquid drains from the riser into the flowline and also builds up at thetemplate end of the flowline. During the rate increase the first slugs were due to the liquiddrained from the riser. The liquid that drained to the template is smeared out along the flowlineto re-establish the equilibrium holdup.

Liquid holdup(vol fraction)

40

Cell number

Time (sets)

Figure 3. Holdup profile during shutdown and start-up


The outlet liquid flowrate predicted by PlAC was converted to a spreadsheet for further pro-cessing by a slug catcher simulation program called PLACSEP, which is available on the PC.This program converts the liquid volume flowrate into the level fluctuations in a cylindrical slugcatcher vessel. Figure 4 shows the level for a 50 bbl vessel (5ft dia x 15ft long) where the liquidpump-out rate is fixed at 25 mbd. In one case there is no control on the minimum level in thevessel, and in the other a minimum level of 20% is imposed.

1.2

0 :

- 20% min holdup

--1-1 Nocon t ro l

8000 12000

1.6

1.4

1.2

Height ’

lrn) 0.8

0.6

0.4

0.2

20% min holdup

No control

16000

Time (s)

i’ I I;:: :I ;II ,I :

.I 1

20000 24000

Figure 4. Slug catcher level fluctuations with and without level control

1 (Swamped)

o-r6000 10000 14000 16000

Time (Seconds)

Figure 5. Effect of pump-out rate on slug catcher level

Pump-out rate

- 25 mbd. . . . . . . 12 mbd

I I

IO 22000 24000



The result is also plotted in Figure 5 for a pump-out rate of 12 mbd and shows that the vesselwould be swamped, even without low level control. Hence it is not possible to use a 50 bblslug catcher if the liquid pumpout rate remains fixed at the maximum subsea production rate.

It is possible by iteration to determine the vessel size vs pump-out rate relationship that justhandles the inlet flow transient. This is shown in Figure 6 and illustrates that if the pump-outrate is to remain at the normal subsea production rate of 12 mbd, then a 275 bbl slug catcherwould be required, but if the spare capacity of the platform process is utilised to give a pump-out rate of 25 mbd, then the required slug catcher volume is around 40 bbls.

350 11

II II II II II II00 50005000 1000010000 1500015000 2000020000 2500025000 3000030000

Minimum pump-out rate (bbl/d)

Ratio of L : D taken as 4 : 1

Figure 6. Predicted slug catcher volume vs pump-out rate relationship

PL4C is now capable of simulating the slug catcher as well as the pipeline and can include rela-tively complex control functions. However it would be very time consuming to use PLAC in thisway to produce the above relationship. It is hence recommended that the PLACSEP programbe used for this purpose. If it is expected that the slug catcher control set-up will influence thepipeline transient, then the PLACSEP program should be used to guess the initial vessel sizeand pumpout rate, and a full simulation used to finalise the design. In the future it is expectedthat PL4C will be interfaced with a process plant dynamic model so that the full pipeline andprocess dynamics may be integrated. This facility is already available via the OLGA/D-SPICEinterface.


Example 7.2 - Simple hand calculation method applied toPompano example

The simple hand calculation method outlined in Section 7.3.1 can also be used to estimate therequired vessel size and pump-out rate for the Pompano restart transient that was simulated byPlAC in Example 7.1.

The starting point for the hand calculation requires the steady state holdup in the flowline at theinitial and final production rates. Figure 1 shows the liquid holdup vs flowrate predicted by MUL-TIFLO for the 6”, 4 mile long Pompano subsea flowline.

1000

800770

cEeE 600rJ2es 4703

400

200

0

- 0% water cut holdup- - - 50% water cut holdup- m - 0% water pig slug- - 50% water pig slug

‘%’tI \At 1 mbd holdup = 770 bbls

j ‘\ At 12 mbd holdup = 470 bblsI ‘\

’ \ j ‘\..*. volume removed = 300 bbls

J ‘.\\\

I ’‘\

- -\

‘-‘-‘-“““““‘-~-~~~

I ---------) MCI I I I I I

0 2 4 6 8Oil flowrate (mstb/d)

10 12

Figure 1. MULTIFLO holdup predictions for Pompano 6” subsea flowline

It is seen that for an initial rate of 1 mbd the equilibrium holdup is 770 bbls for 0% water cut,reducing to 470 bbls at a final flowrate of 12 mbd.

The residence time is approximated by the equilibrium holdup divided by the liquid flowrate,hence:

Residence time at 1 mbd = 770/1000 = 0.77 daysResidence time at 12 mbd = 470/l 2000 = 0.039 days

The transition flowrate is assumed to occur for a period equal to the residence time at the finalcondition and is equal to the final production rate plus the rate corresponding to the excessholdup swept-out during the period of the transition. Hence:

Transient Flowrate = 12000 + (770-470 / 0.039) = 19.692 mbd


Section 7. Transient Flow HP Multiphase Design Munual

The approximate outlet liquid flowrate history is illustrated in Figure 2 and is compared with thePlAC predictions.

19.6 mbd

12.0 mbd

10000 12000 14000 16000 16000 20000 22000

Time (s)

Figure 2. Approximate outlet liquid flowrate profile and PLAC prediction

The slug catcher volume vs pump-out relationship is hence given by a simple linear functionsince the flowrate during the transition is assumed to be constant, therefore:

Surge volume = 0.039 x (19692 - Qout)

hence:

at 0 mbd,at 12 mbd,at 19.692 mbd,

surge volume = 768 bblssurge volume = 300 bblssurge volume = 0 bbls

The comparison between the simple method and the rigorous PLAC simulation is shown inFigure 3 where it is seen that the agreement is close for pump-out rates similar to the final equi-librium value. However, there are large discrepancies at the extremes for the reasons outlined inSection 7.3.1. For zero surge volume the pump-out rate predicted by PLAC is 51 mbd com-pared to 19.7 mbd assuming a constant transition flowrate, hence the simple method fails totake account of the liquid distribution in the pipeline and consequently underestimates the peakoutlet flowrates.


=E

250 -

E1 200 -Fks 1 5 0 -8

30 loo-

E1 200Fks 1508

30 100

=E

250

350 -

50 -50

00 I I I I I I00 50005000 1000010000 1500015000 2000020000 2500025000 3000030000

Minimum pump-out rate (bbl /d)Minimum pump-out rate (bbl /d)

Figure 3. Comparison of slug catcher relationships derived from PLAC and hand calcula-Figure 3. Comparison of slug catcher relationships derived from PLAC and hand calcula-tionstions

Example 7.3 OLGA simulation of Miller landline shutdown andrestart

The 30” Miller landline normally transports dense phase gas over 18 km from the St Fergus ter-minal to Peterhead power station. The gas undergoes dew point control and heating at StFergus to remain single phase. Under normal operating conditions the gas flowrate is 215mmscfd and leaves St Fergus at 37 deg C and 30 bara, arriving at Peterhead at 30 deg C and25 bara. During shut-down and stat--up there is the possibility of liquid condensation and drop-out which may affect the size of the downstream plant and operating procedures necessary tocontrol the liquid outflow. The combination of rate changes and associated cool-down givesrise to a more complicated transient analysis than previously discussed.

In such a transient the amount of liquid formed is a function of the richness of the input gascomposition, the temperature gradient along the pipeline, Joule-Thompson effects, transientcool-down, the topography, and pressure drops. Some of the interactions between theseeffects can be complicated. For example, the pressure drop can cause retrograde condensa-tion which can give rise to higher pressure drops due to interphase friction. This can lead to J-Tcooling which produces more liquid dropout, and so on. Once liquid is produced it may taketime to drain into dips, and may produce slugs. On start-up the the liquid can re-evaporate inthe pipeline before reaching the outlet.

The OLGA code was used to simulate the effects caused by the shut-down and consequentcooling of the Miller landline. The start-up was also modelled in order to study the complextransient multiphase effects taking place throughout the pipeline system as a result of the intro-duction of warm gas into the cool environment. This study formed part of the design stage ofthe pipeline project and provided a valuable insight into the sizing requirements for slug-catch-ing equipment at the power station delivery point.

The transient simulations were performed in three stages:

1. Run to steady state at 215 mmscfd with an inlet pressure of 30 bara and an inlettemperature of 37°C.



2. Shut-down with the outlet closed and the line allowed to pack up to 35 bara throughout. The line is subsequently allowed to cool for 48 hours.

3. Start-up the pipeline with inflow and outflow of 60 mmscfd and remain constant for13 hours. The inlet flowing gas temperature was the order of 80°C. Finally, theflowrate is ramped to 215 mmscfd with an inlet temperature of 37°C and the simula-tion continued until the outlet temperature exceeds 30°C.

The results of the simulation are best described with reference to the pipeline profile and the rel-evant cell numbers illustrated in Figure 1.

80 -

60 -Y-coordinate

(m) 40-

10000

X-coordinate (m)

Figure 1. Miller landline profile

Steady state results gave pipeline outlet conditions of 27 bara and 30.4%. During the shut-down the pressure quickly packs up to 35 bara and the temperature gradually drops, taking 48hours to drop to 6 Deg C at the outlet. This is illustrated in Figure 2. The ground temperature is-3°C. During the shut-down the liquid condenses and drains into the dips giving rise to anincrease in the pipeline liquid content, which still rises after 48 hours as a result of the cooldown(Figure 3).


Fluidtemperature(Degree C)

- Inlet.----- Set 3 6----- Outlet

II II II100000100000 150000150000 200000200000

Time (Seconds)

Figure 2. Variation of shut-in temperature with time

40

30 1Totalliquid

contentin the

system

(M3)

100000100000

Time (Seconds)

Figure 3. Pipeline liquid content during shut-in

In the first phase of the start-up the gas flowrate is ramped upto 60 mmscfd with an inlet tem-perature of 80°C. This has the effect of reducing the downstream pressure as the hydraulic gra-dient is established, and warming the fluid temperature. The warm-up is seen to be slow as theoutlet gas temperature has only risen by a few degrees after 13 hours (Figure 4). The slowwarm-up and flowrate increase produces liquid sweepout at some of the dips. However the liq-uid fills subsequent dips and does not exit the pipeline, hence there is still a net increase in thepipeline liquid content during the first phase of the start-up (Figure 5).

-


Section 7. Transient Flow BP Multiphase Design Manuul

- Inlet. . . . . . . S,.t 36--I- Outlet

80

1

Fluidtemperature(Degree C)

-20 [ 1 I0 10000 20000 30000 40000 50000

Time (Seconds)

Figure 4. Temperature during first start-up phase

-- I0 10600 20600 30600 40600 50000Time (Seconds)

Figure 5. Pipeline liquid content during first start-up phase

In the second start-up phase the gas rate is increased to 215 mmscfd with the gas flowing in at38°C and 35 bara. The further increase in the flowrate results in a higher pressure gradient andhence a decrease in the outlet pressure. The increase in the cold gas inflow rate causes thewarm front to accelerate through the pipeline ( Figure 6) causing liquid to be flashed off, andsweeps out the residual liquid from the system (Figure 7). As a result the total liquid content ofthe system drops sharply removing around 110 m3 (690 bbls) of liquid from the pipeline (Figure8), however around 30% of this liquid is evaporated and the remainder flows to the slug catch-er.


80

Fluidtemperature 60(Degree C)

40

20

04c

- inlet------ Set 36--I- Outlet

I I I I I 1

100 50000 60000 70000 80000 90000 100000Time (Seconds)

Figure 6. Temperature variation during second start-up phase

HoldupC-J

0.10

0.08

0.06

0.04

0.02

0.00

-0.02

- Inlet------ Set 36

:::::::

40000 50000 60000 70000 80000Time (seconds)

90000 100000

Figure 7. Liquid holdup during second start-up phase


Section 7. Transient Flow BP Multiphase Design M6ln~hd

Totalliquid

contentin the

system(M3)(M3)

1.50-1.50-

loo-loo- 11

50-50-

O-O-

-50 -50 II II II II II II II4000040000 5000050000 6000060000 7000070000 8000080000 9000090000 100000100000

Time (seconds)

Figure 8. Pipeline liquid content during second start-up phase

Example 7.4 PLAC analysis of severe shgging in a catenary riser

A comparison has been made between PLAC and some experimental results of severe slug-ging in a Catenaty riser. The experimental data was taken by BHRA in 1990 (Reference 1) andinvolves holdup, pressure, and velocity measurements in a 2” diameter, 108 ft high catenaryriser model using air and water as the test fluids. A 200 ft length of 2” downhill inclined line wasused before the riser and an air buffer vessel of 0.126 m3 was used in the air supply line tomodel a larger pipeline length. The test rig is illustrated in Figure 1.

PLAC simulations have been performed by XFE for a test condition just in the severe sluggingregion corresponding to gas and liquid superficial velocities of 2.3 m/s and 0.33 m/s respective-ly. The measured pressure at various points in the riser are shown in Figure 2 where the toptraces are for the transducers in the inclined flowline and at the base of the riser. The experi-mental severe slugging cycle time is 183 seconds.

The test rig was modelled in PMC as a TEE component with 86 cells, in which a 31 m horizon-tal section of pipe was used to simulate the air buffer vessel. The liquid is introduced into theside arm of the TEE, which is located at the end of the horizontal section. The topography isshown in Figure 3. The PLAC simulations begin with an empty pipe and hence some time isrequired for the severe slugging cycle to be established. Figure 4 shows the pressure in theinclined flowline which is in good agreement with the measured data of Figure 2.

76 Issue 2 2 - December 19941994

33m

w Gamma densitometerl Pressure transducerq Slug progression probe+ Electra magnetic flowmeter

Air and watermixture

/

2” declinedpipeline

w

P 15m

I-

26m

Figure 1. Schematic of experimental riser facility

slug Bubble Gas

Slug build up production penetration blowdow”

r I I I- I

0 1 6 3 2 5 1 6 7 6 3 1 1 0 1 2 6 1 4 2 1 5 4 1 7 0 1 8 6

T i m e ( s )

Figure 2. Pressure traces during one severe slugging cycle


Section 7. Transient Flow BP Multiphase Design Munud

The test rig was modelled in PLAC as a TEE component with 86 cells, in which a 31 m horizon-tal section of pipe was used to simulate the air buffer vessel. The liquid is introduced into theside arm of the TEE, which is located at the end of the horizontal section. The topography isshown in Figure 3. The PLAC simulations begin with an empty pipe and hence some time isrequired for the severe slugging cycle to be established. Figure 4 shows the pressure in theinclined flowline which is in good agreement with the measured data of Figure 2.

3030

2525

2020

Es

1515.-iii5 10w

55

-5 -5 / I I I I I I I I-40-40 -20-20 0 2020 4040 6060 8080 100

Distance (m)

Figure 3. Experimental rig topography used in PLAC

2.52.5

2.02.0

0

300300 340340 380380 420420 460460 500500

Time (s)

Figure 4. Pressure in inclined flowline predicted by PIAC.

Figures 5 to 7 illustrate the holdup and velocity fluctuations throughout the riser. Analysis of thePLAC results enables the following comparison to be made:


Cycle timeSlug build-up timeSlug production timeBubble penetration timeGas blowdown timeSlug lengthMaximum pressure at riser baseSlug tail exit velocity

Experimental measurements PLAC predictions

183s124s24 s14s21 s59 m47.5 psig3.5 rtlls

172s127s11 s9 s25 s57 m48 psig4.5 m/s

From the above comparison it is seen that in most cases the PL4C predictions are in goodagreement with the measured values apart from the slug production time. This time is howeverrelatively short and is difficult to accurately determine from the plots. The experimental condi-tions were not tabulated by BHR4 and hence the estimation of the flowing velocities could alsobe subject to some error. However, the predictions indicate that the flowing conditions in theriser are close to the boundary for true severe slugging as the riser is only full of liquid for a shortperiod of time before the gas pressure is sufficient to eject the slug. Figure 8 shows the testcase point on the experimental flow pattern map and confirms that the PLAC predictions arequalitatively correct.

Figure 5. Holdup variation during severe slugging


Section 7. Transient Flow BP Multipbase Design Manual

VaP vel (m/s)

30302525202015151010

Cell-face number &” o ..

Figure 6. Vapour velocity variation

Liq vel (m/s)

10

88

66

44

22

00

-2-2

-4-4

10001000

)r)r

8o 6~~~400500 Time (sets)

Cell-face number 0 1000 100

Figure 7. Liquid velocity variation -


0.1

1 1 li:Test point

11-T1 11 1 11

/


/

1 Severe slugging 1a Severe slugging la

2 Severe slugging 23 Severe slugging 3s Slug Flow

I

0.1I I

1 10

Gas velocity number

I

100

b Bubble Flow

Figure 8. Flow regime observations from catenaty riser experiments

References

1. G Henday, V Tin, and Z El-Oun “Severe Slugging in Flexible Risers-Interim Report”. MPE048, Jan 1990, Prepared by the British Hydromechanics Research Association.

Example 7.5 - Analysis of hilly terrain pipeline loop for Cusianaphase 1.

The Phase 1 Cusiana development contractor performed a study in October 1992 to determinethe configuration of the production flowlines that were required to transport multiphase wellproduction to the Central Processing Facilities. The flowrates were expected to increase asmore wells were drilled in the future, this was accommodated in the design by extensive loop-ing of the pipelines in order to increase the capacity. The pipe sizes were mainly dictated byflowing velocity considerations to avoid erosion limits, however in the study these were incor-rectly calculated as half the required value, hence the flowlines were significantly oversized. TheMultiphase Flow Group in XFE were commissioned to investigate the proposed multiphasegathering system and to investigate the potential multiphase design issues.

The philosophy of looping was investigated by considering a 20” pipeline loop that was pro-posed between the original well pad 504 and the Cusiana CPF. The topography of this sectionis shown in Figure 1 and illustrates the large elevation changes as the pipeline negotiates thehills, the Cusiana river, and the approach to the CPF which is located on a mesa. Both pipesare assumed to have identical profiles. The analysis of the hydraulic operation of the pipelineloop was conducted using MULTIFLO simulations to determine the possible flowing solutionsbased on a steady state analysis.



ElevationElevation(Feet)(Feet)

1100 1100 !! II II !! II II II II II7000070000 7500075000 8000080000 8500085000 9000090000 9500095000 100000100000 105000105000 110000110000

Distance (Feet)Distance (Feet)

Figure 1. Topography of 20” looped pipeline

The table below shows the calculated gas-oil ratios for a 70 mbd oil flowrate with a producinggas-oil ratio of 2000 scfktbo. At the inlet conditions to the loop of 600 psia and 175°F the cal-culated solution gas-oil ratio is 180.81 scfktbo. (Y is the fraction of the inlet liquid flowrate in lineA and 6 is the fraction of the inlet free gas flowrate in line A.

8

0.00.10.20.40.60.80.91 .oQoa

0.1 0.2181 181

2000 10903819 20007458 3819

11096 563814734 747816554 836718373 9277

7 14

Liquid fraction CY0.4 0.6 0.8

181 181 181636 484 408

1090 787 6362000 1394 10902910 2000 15453819 2606 20004274 2910 22274729 3212 2455

28 42 56

0.9 1 .o181 181383 363585 545989 908

1394 12721798 16362000 18182202 2000

63 70

Effective GOR’s for use in two-phase simulations

The MULTIFLO predicted pressure drops for each gas and liquid flowrate combination areshown in the table below where the pressure drops are calculated for a fixed value of 6 at eachvalue of (Y and hence GOR and oil flowrate Qoa.


8

0.00.10.20.40.60.80.91 .o

0.1 0.2163 162118 130

91 10168 8365 7868 8271 8675 91

Liquid fraction cx0.4 0.6 0.8

163 164 167123 109 122

95 96 11275 94 11480 104 12792 118 14499 127 154

108 137 165

0.9 1 .o169 171129 135120 128124 134138 149157 170168 181179 193

Overall two-phase pressure drops for line A

The data can be plotted as pressure drop vs (Y for each value of 3. If the line sizes or topogra-phies of the loop are different the calculations must also be repeated for line B. However if bothlines are the same the results will be symmetrical ie CY = 0.9 and f3 = 0.9 for line A correspondsto OL = 0.1 and I3 = 0.1 for line B. The result for the case considered is shown in Figures 2 and 3,where the locus of possible solutions is shown in Figure 3.

.-

200

180

160

140

Pipelinepressure

drop 120

WI)

80

60

40 + I I I I I I

0 20 40 60 80 100 120

Liquid fraction to leg

- Gas fraction to leg 0% a*-“- Gas fraction to leg 60%~~~~~~*~~~~~** Gas fraction to leg 10% --I- Gas fraction to leg 80%....... Gas fraction to leg 20% - Gas fraction to leg 90%.-s Gas fraction to leg 40% - Gas fraction to leg 100%

Flowrate = 70 mbd Inlet temperature =175 f Outlet pressure = 500 psig

Figure 2. Solution for line A



180

160

140

Pipelinepressure

drop 120

(PSI)

80

60

40 I

20I I I I I

40 60 80 100 120

Liquid fraction to leg

d

- Gas fraction to leg 0% ~~-~*- Gas fraction to leg 60%~~~~~~~~~~~~~~ Gas fraction to leg 10% X--B- Gas fraction to leg 80%.*..*.. Gas fraction to leg 20% - Gas fraction to leg 90%I - - Gas fraction to leg 40% - Gas fraction to leg 100%

Flowrate = 70 mbd Inlet temperature =175 f Outlet pressure = 500 psig

Figure 3. Combined solutions

The figures illustrate that at 70 mbd no zero flow or manometer effect can exist since 0% liquidfraction is not a solution. The nature of the tee junction at the loop is such that the liquid maypreferentially take the run and this could give a 100% free gas split into the other leg with 22%of the liquid flowing in the run. Note that between 40% and 60% liquid split to the run severalsolutions are possible with different overall pressure drops. The main problem is the liquidholdup variations in each leg if the flow switches. The holdup at 0% gas split to the run is 9377bbls and the flowrate is 15.4 mbd. In the side leg the corresponding holdup is 2264 bbls givinga 7113 bbl slug if the flow in the loop switches. The possibility of unstable operation was part ofthe reason for recommending that the Cusiana multiphase flowlines should not be looped.

For the example above the sum of the uphill elevations is 504 ft, which gives rise to a hydrostat-ic head of 163 psi. Figure 4 indicates that manometer effects are possible when the inletflowrate is reduced below around 60 mbd since the flowing pressure drop could be balancedby a static column of liquid in the other leg.


200

Pipelinepressure 150drop (psi)

100

50 I I

20 40 fI I

8 0 100

Oil flowrate (MSTB/D)

Figure 4. Pressure drop characteristic for total flow in the one leg

Example 7.6 - Using two-phase pressure surge analysis to deter-mine pig-slug loads

The effect of a pig-slug impacting on a closing pressure control valve was studied as part of thedesign of the Rough field slug catcher at Easington. An isometric of the approach pipework tothe slug catcher is shown in Figure 1 where it is seen that the 34” sealine terminates in a 17mriser at ‘A’ then follows horizontally, and finally rises at 45 degrees between ‘D’ and ‘E’ beforeentering the sphere receiver. The PCV is located just upstream of the slug catcher. The PCV isdesigned to control the downstream pressure independently of upstream conditions. Flow con-trol is provided downstream of the gas offtake ‘J’.

It was required to estimate the loads produced by the pig-slug impacting on the PCV if it closedduring pig reception, and hence to determine the need to set the valve fully open during pig-ging. The slug catcher is located 1 Om above ground level and it was required to assess thestructural loads that may be generated by the slug dynamics.

A simple dynamic model of the slug reception process indicated that the slug velocity increasesfrom a initial value of 7m/s to a final velocity of 21 m/s as the tail passes the PCV and the slug isdischarged into the slug catcher. This acceleration is due partly to the hydrostatic head lossand the reducing frictional length of the slug as it is produced.

initial simulations using FLOWMASTER were based on the surge pressures generated by clos-ing the PCV during slug reception. A linear valve closure rate was assumed and the effect of the


Section 7. Transient Flow BP Multiphase Design Mmmd

valve closure time investigated. The results of a single phase surge analysis are shown in Figure2 indicating that surge pressure should not exceed 95 bar, which is well below the maximumallowed pressure of 150 bar. The valve closure times of around 30s are too long to generatesignificant unequilibriated loads in the short piping runs. What was of more concern was theimpact of the slug front on a partially open control valve, which is more difficult to assess.

Offshore / onshore tie-in

/

B Sphere

W Open valveH Shut valve

catcher

Figure 1. Isometric view of approach pipework to the slug catcher

65

Valve closure time (s)

60 1 I I I I I I I I I I0 10 20 30 40 50 60 70 8 0 90 100

Time (ms)

Figure 2. Results of single phase liquid surge analysis

One of the major unknowns is the time over which the slug front impact occurs. The worst caseis to assume a vertical slug front and hence an instantaneous impact. The results of using theprimer component in FLOWMASTER to simulate this is shown in Figure 3 for locations ‘F’, ‘G’,and ‘H’. The slug was in this case assumed to be travelling at 6 m/s through a 20% open con-trol valve. An all liquid slug was assumed.


1.8

0.6

0 100 200 300 400

Distance from catcher (m)

Figure 3. Pressure histories for worst-case impact assumptions

In practice the slug may typically have a front sloping at 30” which will have the effect ofincreasing the impact time. This was also investigated using FLOWMASTER, showing thatincreasing the impact time from 0 to 50 ms has the effect of reducing the peak surge pressurefrom 140 tonnes to 20 tonnes (see Figure 4). At an arrival speed of 6 m/s this is equivalent to aslug front length of one-third of a pipe diameter, and hence is quite feasible. Although it wouldseem likely that long impact times, and hence small loads, should prevail, the dynamic nature ofthe slug front makes it difficult to accurately predict the impact time and the possibility of aninstantaneous impact should be considered.

160 -

Limit imposed by acoustic velocity of liquid

Slug passage during impact

I I I I0 25 50

Impact time (ms)75 100

Figure 4. Influence of impact time on load predictions

The reaction loads due to the unhindered slug passage are more easily calculated. These aredue to the change in the momentum between the liquid slug and the gas and are related to theslug front and tail velocities. Ignoring elevation effects the slug front velocity is constant and


Section 7. Transient Flow BP Multiphase LIesi@ Manual

hence the ‘coming on’ load is fixed. However the slug tail accelerates during slug reception,and hence the largest loads are the ‘coming off’ loads produced by the passage of the slug tail.These loads are shown in Figure 5 and indicates that the maximum load is of the same order asthat produced by a 50ms impact surge. Slug loads are further discussed in Section 11.

- Region of interest

Load due to tail

Load due to head/

100 200 300 400Distance from catcher (m)

Figure 5. Influence of slug passage on dynamic loads

88 Issue 2 - December 1994~_______~~~

-

BP Multiphase Design Munual Section 8. Depressuring /Emergency blowdown

Section 8 Depressuring andemergency blowdownof single andmultiphase pipelines8.1 Definition of terms

8.2 Depressuring of Single and Multiphase Pipelines

Appendix 8.2a Heat Transfer Models

Appendix 8.2b Nomenclature

8.3 Emergency Blowdown of PipesContaining Gas and Liquid

Issue 3 - January 1998

Section 8. Depressuring /Emergency blowdown BP Multiphase Design Munu~l

-


LEFT BLANK

-

2 Issue 3 - January 1998

BP Multiphase Design Munual Section 8. Depressuring / Emergency blowdown

8.1 Definition of the terms:Depressuring, Blowdown and Rupture

In this manual the terms depressuring, blowdown, and rupture have the following meanings:

Depressuring

Depressuring is generally used to refer to the controlled and relatively slow evacuation of apipeline system. Depressuring is usually performed to make the pipeline available for mainte-nance or repair. Depressuring a pipeline will usually take several hours or even days.

The simulation of pipeline depressuring is discussed below in Section 8.2.

Blowdown

Blowdown is generally used to refer to the controlled but rapid evacuation of a pipework orprocess plant vessels. Blowdown is sometimes referred to as emergency depressurisation.

Blowdown is used to minimise the escalation of any incident by reducing the risk of pipework orvessel rupture in a fire. In addition blowdown will reduce the hydrocarbon inventory which couldfeed the fire. Blowdown is normally accomplished in a time of the order of 15-30 minutes.

Blowdown philosophy and modelling is discussed below in Section 8.3

Rupture

Accidents resulting in fluid loss may occur as a result of overpressurisation, material failure andimpact damage. The fluid loss occurring as a result of a rupture will depend on fluid composi-tion, hole size, initial pressure, pipeline characteristics, and mass of fluid available in the pipe tofeed the rupture.

The fluid obviously escapes from a rupture in an uncontrolled manner. Pipeline rupture model-ling is discussed in Section 9 of this manual.

Issue 3 - Januarv 1998 3

Section 8. Depressuring /Emergency blowdown


LEFT BLANK

4 Issue 3 - Januarv 1998

BP Multiphase Design Manual Section 8. Depressuring /Emergency blowdown

8.2 Depressuring of single and multiphase pipeline systems

8.2.1 Introduction

The Model discussed here is used to determine a conservative minimum pipe temperatureexperienced during the controlled depressuring of a pipeline system. The model is applicablefor the relatively slow depressuring of a system over many hours/days. For more rapid depres-suring events the approach discussed here will produce very conservative temperatures i.e. itwill significantly under predict the minimum pipewall temperature. The blowdown modeldiscussed in Section 8.3 below is more appropriate for the rapid emergency depressuringevents which occur over a time scale of 15-30 minutes.

The Pipeline model described in this section is applicable for horizontal or slightly inclined singleand two phase systems, It also predicts liquid entrainment rates and the critical velocitiesrequired to produce slug or annular-mist flow in a riser. The temperature downstream of theletdown valve/orifice is also computed.

8.2.2 Overview

The simulation of the depressuring of a pipeline system is performed in order to determine theminimum temperature experienced, or to calculate the time required to depressure a systemsuch that the temperature does not drop below the minimum specification value.

The purpose of a depressuring calculation is therefore to determine the relationship of pressureand temperature against time for a particular system and depressuring rate.

The calculation has three main elements. These are;

(a) Depressuring Rate

The depressuring rate is a critical aspect. The faster the depressuring the lower the temperaturethat will be experienced.

The venting rate can be fixed (e.g. 100 mmscfd) by using a control valve and a mass flowratemeter. Alternatively, the system can be vented through an orifice. In most cases the flowthrough the orifice is critical and therefore the depressuring rate is independent of the down-stream pressure.

(b) Auto-refrigeration

As the pressure decreases, the temperature of the system will drop due to Joule-Thompsoncooling (auto-refrigeration). The expansion of the gas is normally considered to be isentropic. Asthe gas expands it does work and the energy required to do this work is ‘removed’ from thegas in the form of heat. The auto-refrigeration is not dependent on the duration of the depres-suring, only the physical properties of the fluid.

Issue 3 - Januarv 1998

(c) Heat Input from the Surroundings

To depressure a pipeline system usually requires many hours, and in some cases days. Duringthis period the heat input from the surroundings may be large.

The auto-refrigeration is not time-dependent and is only dependent on the expansion of the gasin the system. The ‘heat from the surroundings’ is time dependent. The longer the depressuri-sation, the greater the influence of the surroundings whilst the auto-refrigeration aspect remains(roughly) constant. Therefore, increasing the depressuring time will result in a higher minimumtemperature.

8.2.3 Depressuring Rate Cahdation

It is usual to depressure a system through an orifice or a valve. The equation for the massflowrate through an orifice is given below (Equationl). The equation for a valve is of a similarform with slightly different constants (not given here).

W=C,KAP, pJTT

(For nomenclature, see the section at end of this document.)

K, the specific heat ratio function is defined by the following equation;

K=m

(1)

(2)

The equation above is only valid for critical (i.e. sonic) flow. Critical flow only occurs when thefollowing criteria is satisfied (equ.3).

p2 2

1 I(Y/Y’)

-< - (3)

PI Y+l

The occurrence of critical flow greatly simplifies the calculation. If the system is operating undercritical flow then the fiowrate is only dependent on the upstream pressure.

It is normal to consider the depressuring to be complete when the system pressure falls below2 bara. For a downstream pressure of 1 bara, critical flow exists down to pressures of 1.83bara (r = 1.3) and so therefore critical flow is usually assumed for the whole of the depres-suring. Critical flow may not exist at the lower flowrates where the downstream pressure issignificantly greater than atmospheric,

Whilst the pressure ratio satisfies Equation 3, the flowrate (W) is independent of the down-stream pressure.


BP Multiphase Des&n Manual Section 8. Depressuring /Emergency blowdown

8.2.4 Heat Transfer

(a) Factors Affecting the Minimum Temperature

The temperature of the fluid is determined by the effect of cooling due to the Joule-Thompsonexpansion of the gas and warming by the influence of the surroundings. The conservatism ofthe equation is described below;

Isentropic Expansion.

Isentropic expansion means that the gas does the maximum amount of work as it expands.This means that it loses the maximum amount of energy (i.e. heat) and therefore results in theprediction of the minimum possible (i.e. conservative) temperature.

Warming Influence of the Surroundings.

When the gas temperature drops below the ambient temperature the pipeline is warmed by thesurroundings. To calculate a conservative (i.e. low) minimum temperature the influence of thesurroundings must be underestimated.

By overstating the cooling effects of expansion and understating the heating influence of thesurroundings, the calculation of a conservative minimum temperature is ensured.

(b) Internal Heat Transfer Coefficient

The calculation of an accurate internal heat transfer coefficient is important since it greatly influ-ences the prediction of the minimum temperature. The method used must calculate a conserv-ative (i.e. low) heat transfer coefficient for at least the period that the system is at its minimumtemperature.

The correlation used to determine the heat transfer coefficient will depend on the type ofsystem and the method of depressuring. Details of the equations used to calculate the internalheat transfer coefficient are given in Appendix 8.2.A.

(c) External Heat Transfer Coefficient

The external medium has a significant effect on the calculated minimum fluid/wall temperature.If the pipe is located in the air (e.g. topsides of a platform) then the overall heat transfer coeffi-cient will be low, giving rise to a relatively low minimum wall temperature. If the pipe is located inthe sea the overall heat transfer coefficient will be higher and the minimum wall temperature willaccordingly be higher (assuming, for this comparison, that the air and sea temperatures areequal)

(d) Minimum Tube Wall Temperature

The objective of most depressuring studies is the calculation of a minimum pipe wall tempera-ture, or the calculation of a depressuring rate which prevents the system from falling below theminimum tube wall temperature.

In a gas-only system the heat transfer rate between the gas and the pipe wall will be low, andconsequently the temperature difference between the fluid and the tube wall will be relatively

Issue 3 - January 1998 7


high. In order to be conservative, it is usual to assume that the fluid temperature and the tubewall temperature are identical throughout the depressuring. This assumption will always result inthe calculation of a conservative (i.e. low) minimum tube wall temperature since it ignores theexistence of a temperature profile between the fluid and the wall.

In a two-phase system, the heat transfer rate between the liquid and the pipe wall will be signifi-cantly higher than the heat transfer rate between the gas and the pipe wall. It is reasonable toassume that where pools of liquid occur (in dips etc.) the high heat transfer rate will result in thepipe wall cooling to the same temperature as the fluid. However the effect of the fluid beingwarmed by the pipe wall is ignored and hence a conservatively low pipe wall temperature ispredicted.

82.5 Simultaneous ‘Riser’ and ‘Topsides’ Depressuring

In most studies the ‘riser’ and ‘topsides’ pipework will be depressured simultaneously. Themajority of the length of the riser will be immersed in the sea and have the benefit of the sea asa heat source. Riser pipework is generally much simpler than the Topsides pipework.

The topsides pipework is more complicated (e.g. a large number of branches etc.), and locatedin the air. The poor heat transfer properties of air coupled with the lower minimum ambienttemperature mean that the topsides pipework will drop to a lower temperature relative to theriser. Because of this difference in heat transfer rates the topsides and the riser should bemodelled separately.

8.2.6 Critical Velocity for Slugging

Slugging should be avoided where possible. To avoid this the depressuring rate should notexceed the critical velocity for slugging. The angle of the pipe (to the horizontal) is used tocalculate the conditions required for slugging.

8.2.7 Flow Regime in Riser

Annular mist flow is unlikely to occur in the riser. The actual velocities experienced in the riserare usually approximately an order of magnitude less than that required to produce annular-mistflow.

8.2.8 Conditions Downstream of Letdown Valve

The temperature downstream of the let-down valve will invariably be significantly colder thanthat experienced in the topsides or riser pipework. Temperature specifications of the down-stream pipework are not usually a problem since stainless steel is usually the material ofconstruction. Liquid drop-out may occur due to the low fluid temperature, and this will need tobe accommodated in a liquid KO vessel.

-


i

BP Multiphase Design Manual Section 8. Depressuring / Emergency blowdown

Appendix 8.2a

Heat Transfer Equations

The calculation of an accurate internal heat transfer coefficient is important since it greatly influ-ences the prediction of the minimum temperature. The method used must calculate a conservative (i.e. low) heat transfer coefficient for at least the period that the system is at its minimumtemperature.

The correlation used to determine the heat transfer coefficient will depend on the type ofsystem and the method of depressuring.

The internal heat transfer coefficient is calculated using one of the following equations:

Stream Line Flow (Re C 2100)

If the flow is laminar (Reynolds number less than 2100) then the following equation should beused:

hi = 1.7 [ ( di%J( c;Jj(+] ‘I3

[“Process Heat Transfer”. D.Q. Kern. equ. (6.1). page 1031

As the above equation and those following will usually be applied to non-viscous gases theviscosity correction factor is ignored.

Correction For Natural Convection

For very low flowrates natural convection (rather than forced) convection can become impor-tant. The forced convection coefficient (hi) can be corrected by multiplying by

2.25 [l +O.Ol (Gr,“3)]T=

log,, Re

[“Process Heat Transfer”. D.Q. Kern. Equ (10.4). page 2061

Where Gra is the Grashof number at the average conditions, given by:

Gra =

[“Process Heat Transfer”. DLL Kern. Table 3.2. page 371

(4)

(5)

(6)

This factor should be ignored if less than 1.The Grashof number is difficult to calculate since the temperature difference between the bulk


Section 8. Depressuring / Emergency blowdown

fluid and the tube wall (AT) has to be determined iteratively. It can be done by guessing a walltemperature and iterating such that the heat flux across the internal film is the same as the heatflux through the pipe and the insulation.

If this correction is ignored, then the internal heat transfer coefficient will be under-predictedand a conservative (i.e. low) heat transfer coefficient will be determined.

Turbulent Flow

If the Reynolds number is greater than 2100, then the following should be used

(7)

[“Chemical Engineering. Volume 1 ‘I. Coulson and Richardson. Equ (7.50). page 1911 -

(Note; This is the same method as MULTIFLO, BP’s in-house two-phase pipeline modellingprogram.)

Alternatively, Equation 7 can be re-written as below:

hi = 0.023(k,)0.6 (di)-O.* (/A,,,, pJO.8 (/+,,)-“.4 (CJO.4

m denotes a mixture property, e.g. for viscosity.

hll = PI A + F, (1 -N (9)

The velocity at which the heat transfer coefficient is to be calculated should be 75% of the exitvelocity. This allows for the fact that there is a velocity profile through the system. If possible thevelocity should also be based on the average velocity for the time step.

-


Section 8. Depresswing /Emergency bZowdown

Appendix82b

Nomendature

A

cciCd,Pm

Cl

Gr8h,K

kmL

VWp,p2RReT

MULTIFLO

EXTMEDINSLTN

= Orifice/valve area for venting= Coefficient of Discharge= Mixture heat capacity= Internal pipe diameter= Gravitational constant= Grashof number= Internal heat transfer coefficient= Function of specific heat ratio= Mixture thermal conductivity= Length of System= Molecular Weight= Upstream Pressure= Downstream Pressure= Gas constant= Reynolds’ number= Temperature= Mixture velocity= Mass flow rate= Thermal expansivity= Ratio of Specific Heats= Volume fraction of liquid= Liquid viscosity= Mixture viscosity= Vapour viscosity= Mixture density= Free Convection Coefficient= Gas compressibility

= External medium h.t.c.= Insulation h.t.c

PIPEWL = Pipewall h.t.c

m*

I-1 I[J/kg Kl[ml[9.814 m/s?]

L-1[W/m*.K][kg-mo1°.5.K0,5J-0.5]

Wm.Kl[ml[kg/kg-mole][N/m*][N/m*][8314 J.Kgmole-l .K-l]

I-lWIk-WWsl[l/K]L-1i-1[Ns/m2][Ns/m2][Ns/m2]

k/m31L-1i-1

[W/m*.Kjv/m*.K][W/m2.K]

issue 3 - January 1998 I I


-


LEFT BLANK



8.3 Emergency blowdown of pipes containing gas and liquid

8.3.1 Background

This section of the manual is concerned with the simulation and philosophy of emergency blow-down (i.e. rapid depressurisation) of pipes and vessels containing gas and liquid. Conditionsexperienced during blowdown generally set the minimum design temperature of those pipesand vessels subject to emergency blowdown.

A review of blowdown issues concluded that the only program that could accurately model theblowdown process was developed by Saville and Richardson of Imperial College London. Theyhave been working in the area for many years developing both the experimental facilities toprovide accurate data, and the theoretical models to simulate the blowdown process.

Saville and Richardson have been contracted on many occasions by BPX to carry out blow-down analysis of process plant and pipework. However, BP identified a need to have a simpli-fied blowdown model available to assist design teams in carrying out sensitivity studies and indeveloping optimum blowdown procedures. A blowdown module has consequently beendeveloped within the GENESIS flowsheeting package using a simplified version of the approachtaken by Saville and Richardson. This is discussed below in 8.3.2.

8.3.2 Blowdown Philosophy

It is important that a blowdown philosophy document is developed in the early stages of anydesign, defining not only the relevant cases but also the start conditions from which the blow-down occurs.

(a) Code Requirements

API RP 521 provides the general guidelines for determining the need for emergency blowdown.There are, in fact, no mandatory blowdown requirements. However, emergency blowdown hasbecome accepted in the industry as being necessary for all offshore installations, other than forsome unmanned platforms. The BP Code of Practice CP 37 clearly defines this philosophy.

CP 37 sets out the 3 basic reasons for providing emergency blowdown facilities as:

1. To reduce the risk of vessel or pipeline rupture in a fire.2. To minimise the fuel inventory which could supply a fire.3. To minimise the uncontrolled release of flammable or toxic gas.

Emergency blowdown facilities are not required for subsea pipelines. The platform topsidessection of the pipeline, including pig launcher/receiver and all pipework on the platform side ofthe ESDV should be blown down through the platform blowdown system. For pipelines fittedwith subsea isolation valves (SSIVs) there is also a requirement to provide for emergency blow-down of the riser and subsea pipework situated between the SSIV and the top of the riser.

(b) Effects Of Blowdown

Blowdown of the contents of a vessel, pipework, or sections of a pipeline from high pressure



results in auto-refrigeration of the contained fluid due to expansion of the gas and evaporationof the liquid. The resultant fall in the temperature of the pipe or vessel wall metal can result inthe need for costly materials to avoid brittle fracture. In addition low fluid temperatures canresult in the formation of hydrates, ice, or even solid CO,, and these can lead to blockages. Thedetermination of minimum fluid and metal temperatures is discussed below in Section 8.3.3.

(c) Blowdown Times and Start Conditions

Blowdown must be at such a rate that equipment subject to high temperature does not fail. APIRP 521 sets out the basic requirements for blowdown times and these are further detailed inCP 37. In general blowdown to 50% of design pressure in 15 minutes must be achieved,assuming wall thicknesses of 25 mm and greater. For vessels designed to BS5500, reductionof pressure to 6.9 barg in 15 minutes must be achieved unless the rate can be related to thedecrease in strength of the vessel.

Identification of realistic worst case conditions at the start of a blowdown is of great importancein establishing temperature and pressures experienced during the blowdown process. Worstcase conditions would normally involve starting the blowdown from the maximum pressure andminimum ambient temperature. However, it is also quite possible to have start conditions wherethe fluid temperature is below ambient.

The blowdown simulation of Ravenspurn South was a case where the fluid start temperaturewas below ambient. During platform start-up cold wellhead fluids flash across the choke andcool to temperatures well below ambient. A shutdown of the plant at this time leaves fluidsshut-in at a high pressure and low temperature. These conditions had to be taken as the worstcase starting conditions for the blowdown modelling.

S-3.3 The GENESIS Blowdown Model

A blowdown module has been developed within the GENESIS flowsheeting package using asimplified version of the approach taken by Saville and Richardson. The GENESIS blowdownmodule contains all the most important aspects of the IC model, but it is only capable of simulating a single pipe or vessel, i.e. it cannot be used to model the blowdown of an integratedprocess plant.

In summary, the model uses the following approach:

a. The Vessel is divided into two zones. The top zone contains all the vapour together with anysuspended liquid droplets.

The bottom zone contains the liquid pool which has dropped out of the top zone. The bottomzone appears whenever liquid is formed and is eliminated if there is no liquid pool at the bottomof the vessel.

b. Each zone is assumed to be well mixed and separately exchanges heat with the vessel wallwith which it is in contact.

Heat transfer to the gas zone is assumed to be by natural convection. Studies carried out atSunbury have validated the expressions used to model this process.

14 issue 3 - Januarv 1998


Heat transfer to the liquid zone from the wall is governed by nucleate boiling.

c. Heat transfer to/from the surroundings is not normally significant over the time periodsinvolved in blowdown, and is therefore not considered.

Heat transfer around the walls by conduction is not considered.

d. The model does not consider a suspension of liquid droplets in the vapour zone. All liquidformed is assumed to instantly fall into the bottom zone of the vessel.

Model Validation

Because of the extensive validation work carried out by IC in the course of the development oftheir model, it has been assumed that the IC test cases provide accurate bench mark data withwhich to assess the accuracy of the BP model. The predictions of the BP model were found tobe in reasonably close agreement with those of the IC program.

In all the comparisons performed the BPX model provides lower predictions of wall temperaturethan the IC model. Hence the BPX model is providing more conservative results that the ICmodel.


Section 8. Depressuring / Emergency blowdown


LEFT BLANK


Section 9. Pipeline RupturesModelling

Uncontrolled Rupture of Pipeline andPipework Systems

9.1

9.2

9.3

9.4

Introduction to Basic Principles

The PBREAKZ. program to Model the release ofVolatile Liquids from a Pipeline Rupture

Factors affecting the Transient Analysisof Pipeline Ruptures

Rupture Case Study

References

Issue 2 - December 1994 I

Section 9. Pipeline Ruptures BP Multiphase Design Manual

-


LEFT BLANK

-



ISection 9. Pipeline Ruptures

-

9.1 Introduction to Basic Principles

Pipeline accidents resulting in fluid discharges may occur for many reasons including materialfailure, overpressurisation, and impact damage. If a pipe rupture occurs when a pipeline istransporting a highly volatile liquid, then on depressuring the bubble point can be reached and atwo-phase mixture of gas and liquid released into the environment local to the rupture, as illus-trated in Figure 1.

Single phase Two-phaseregion region

Single phaseregion

Bubble pointconditions

conditionsPbrk Xbrk Tbrk

-2 -IJ

Inlet Environmental Ountletconditions interface conditions conditions

PI, Tm W brk Release flow Pback %ackTback Pout Tout

W,

Figure 1. Conditions in the vicinity of a rupture

Single phase pipelines in the above category include those transporting LNG, LPG, gasolinesand spiked crude oils, all of which can present a significant hazard if accidentally released intothe environment. When pipelines transport a two-phase mixture of gas and liquid a rupture cancause rapid depressuring and critical flow at the break, additionally, with gas/condensate linesthe temperature can drop rapidly due to Joule-Thompson cooling, this can have implications onthe brittle fracture of the pipeline and safety valves which may cool significantly before beingactuated. The analysis of such situations involves complex interactions between the depres-suring hydrodynamics and thermodynamics.

Within the multiphase flow group several tools are available to perform this type of analysisranging from simple specific codes such as PBREAKL to more complex general purpose simu-lators such as PLAC. Which one to use depends on the type of analysis and the accuracyrequired. These simulations are usually needed to support safety assessments including theimpact of oil spills, gas leaks, jet fires, pool fires, and related consequence analysis.

When depressuring a gas the final temperature depends to a large extent on the type of expan-sion process assumed. Minimum final temperatures result from an isentropic expansionprocess which is a reversible adiabatic and frictionless process. If however an irreversible isen-thalpic process is assumed then the friction losses result in a higher final temperature. This isillustrated in Figure 2 which shows the various expansion processes for methane expandingfrom 20 atmospheres and 300 K (27 “C) to atmospheric pressure. The isentropic expansion(line A) results in a final temperature of 140 K (-133 “C) where as the isenthalpic process (line B)



results in a much higher final temperature of 273 K (17 “C). Although the isentropic expansionprocess is often used for a conservative estimate, and is probably appropriate for a full borerupture analysis, some consideration should be given to the configuration of the upstreampipeline as the close proximity pipeline fittings on topside pipework can result in a throttling typeof expansion process which is more isenthalpic. The non-ideality of the throttling process isusually accounted for by an isentropic efficiency term q which is defined as the ratio of theactual enthalpy drop to the isentropic enthalpy drop. For the example in Figure 2 (line C) theisentropic efficiency is given by:

4530-3770‘1= = 62%

4530-3300

It is also seen that for a pure component substance the final temperature is independent of theexpansion process if the final state is two-phase. This would be typical of a refrigeration plantwhere a throttle is used to de-pressure the refrigerant and cause the temperature to drop to thesaturation value associated with the final pressure, hence the same final temperature isachieved regardless of whether the process is isentropic or isenthalpic. In this case the effect ofthe irreversible nature of the throttle is to increase the enthalpy of the fluid entering the cold boxand hence to reduce the change in enthalpy available for refrigeration.

If a multi-component mixture is expanded into the two-phase region, then there is no longer afixed saturation pressure and temperature, since the temperature is a function of the liquidmass fraction. The expansion process is hence important with multi-component fluids.

In view of the difficulty of determining the irreversible nature of the expansion process it isrecommended that first estimates are based on an isentropic process since this is likely to beconservative, but is also expected to be the most accurate when applied to the de-pressuringof large volumes of fluids such as in the case of oil and gas pipelines. By far the most commonrequest from a transient rupture analysis is an estimate of the fluid lost to the environment andthe temperatures of critical components.

The need for safety assessment of the consequences of the release of volatile liquids from arupture in a pipeline prompted the development of the PBREAKL program within BP. This codeis specifically for volatile liquids normally transported single phase, but allows for flashing whendepressured. In addition to this code the multiphase group also has a code called PBREAKGwhich is for the simulating gas pipeline rupture. These codes are limited to the application forwhich they were developed, where as general purpose transient two-phase codes such asPLAC and OLGA can simulate the depressuring of single-phase and multi-phase lines and arerecommended for most cases. PBREAKL has some useful features and is relatively simple touse. Some of these capabilities are outlined below.


BP Multiphase Design Manual Section 9. Pipeline Ruptures

Pressure, atm. ohs

Entholpy, cal /mo le - - -

V o l u m e , c u c m / m o l e - - - - -

l mole 2 16.0429I col = 4.1868 joules

Entropy and entholpy zero forperfect crystal at absolute zeroof temperature

46(

44(

4%(

40(

38(

361

34(

32 34 36 38 40 42 44

tntropy, Cal. (moie)f”K.I

Figure 2. Temperature-entropy diagram for methane

-


I



LEFT BLANK


-


9-2 The PBREAKL program to model the release of volatileliquids from a pipeline rupture

The PBREAKL program predicts the single or two-phase discharge of fluids resulting from arupture of specified size in a pipeline, including the single-phase depacking region to the bubblepoint. The program employs vapour-liquid equilibrium conditions and thermodynamic behaviourof two-phase flowing fluids to predict both critical and sub-critical release rates. The thermodynamic data for PBREAKL is generated using the FlA02 module in the GENESIS package toproduce a physical property look-up table. The critical two-phase flow release predictions areaccomplished using the method of HenryFauske (Ref 1) or Morris (Ref 2), the Morris method isrecommended. For subcritical flow a two-phase multiplier method is applied to a standardsingle phase discharge orifice equation.

PBREAKL can model the effects of a simple pipeline topography with import and export risersand an isolation valve. A schematic of the physical geometry is shown in Figure 3.

InletElevation

Ein

A ~ in

Distance to Valve ABC

Distance to Rupture ABCD

Length of Export Riser AB

Length of Import Riser EFLength of Subsea Line BCDE

OutletElevation

Rhpture DoDiameter

i -

Wall eRoughness

Figure 3. Schematic of pipeline system

PBREAKL has an interactive menu input where the physical dimensions are entered at the pipegeometric details screen. In addition the break diameter, the location of the rupture point andthe back pressure are entered at the pipe break details screen. The flowing conditions aredefined, such as inlet and outlet pressures, liquid flowrate, viscosity, and the temperature overthe pipeline which is assumed to be the average line temperature. The time to valve closure (theclosure is assumed to be instantaneous) and other program run time data is also input. Beforethe program can be run GENESIS phase equilibrium component data and flash data arerequired.

Prior to a typical rupture the fluid is usually said to to be “packed” within the pipeline at elevatedpressures above the bubble point pressure, as a result of pipe wall flexibility and fluid



compressibility. When the pipeline is ruptured the initial release is high as the pipeline depacksto the bubble point at the rupture location. When the flow flashes the initial flow is usuallychoked or critical and the rate of efflux is reduced. As the pressure drops within the pipeline theflow can eventually become sub-critical and the simulation stops if the release rate from therupture equals the inflow rate.

The depacking calculation in PBREAKL is controlled by an expansion factor and an expansiontime constant, which alter the wave propogation effects. These are most significant during theinitial release, and as a result several types of depacking calculation are possible. These areoutlined in Figure 4.

A description of each depacking method is as follows:

Method A.

This situation occurs when the leak rate reaches a steady state which is less than the inflo rateto the pipeline.

Method Bl.

Here the initial leak rate evaluated at the bubble point pressure is higher than the steady inflowto the pipeline, hence the pressure drops almost instantaneously to the bubble point pressure.

Method B2.

The initial release rate exceeds the steady pipeline inflow rate but is less that the maximum rateat which fluid “enters” the two-phase region. Under these conditions depacking can take a finitetime to occur.

PBREAKL indicates the type of flow solution and depacking calculation which is used and tabu-lates rupture gas and liquid rates, the location of the two-phase interface, the total massremoved from the rupture, the rupture pressure and physical properties. All of this data can beplotted graphically.

Some examples of PBREAKL program predictions are given in Figures 5 to 8 for the case of an8” nominal bore pipeline carrying NGL fluid at 10 C. The pipeline is 10 Km long and the initialpressure is 30 bar at a feed rate of 25 kg/s. Note the different depacking characteristics inFigure 6.

-


BP Multiphase Design Manual Section 9. Ptpeline Ruptures

Method A: Small break release

~~-

Method Bl: Instantaneous depacking

Time (s)

A

27.cnc5 Method 82: Finite depacking timeal3

312

IIbb

toto t Bub Time (s)

Time of pipe rupture = t, = 0.0

Time at which bubble pressure reached = tB&

Figure 4. Types of discharge time history in PBREAKL



300

250

? 2 0 0I?aH 150

8d2 100

50

2 5 0

I I I I I I

0 1 0 2 0 3 0 4 0 5 0 6 0

Time (s)

I1000

Time (s)

I I1500 2000

- 1 bar - - - 5 bar --..-. IO bar

Figure 5 Effect of back pressure

2 0 0 0 -

77j 1 5 0 0 -

0 1Y

fj lOOO-:

II

Ii,

2 500- :

:z.-~‘,““““-.--..-...........~..~~,---------------.0 I I I I I I

0 1 0 2 0 30 4 0 5 0 6 0

Time (s)

1000

Time (s)

1500 2000

- 0.05 m - - - 0.102 m -mm--- 0.203 m

Figure 6 Effect of break diameter


BP Multiphase Design ManualSection 9. Pipeline Ruptures

a1 4 0

2 0

---...._--.I...-

0 , I I I I I I

---1. 01 0 2 0 3 0 4 0 5 0 60120 -....... -----.............__

T i m e (S)-

. . . . . . . -.. . . . . . .100 -6 ‘\ ~ -- --aB

--- ----- 8 0 - ---w-s-22

------------ ---.

E8 6 0 -1$

40 -

20 -

0 I II I I I I

0 10 20 30 40 50 60 70

Time (s)

- -1O’C - - - 1O’C . . . . . . 3O’C

Figure 7 Effect of inlet temperature

500

1

I I I 1 I I I0 1 0 2 0 30 4 0 5 0 6 0 7 0

T i m e (s)

I1000

Time (s)

- Fauske - - - - Morris

Figure 8 Effect of choked flow correlation-

I IIssue 2 - December 1994

Section 9. Plpelhe Ruptures BP Multiphase Design Manual

Experience with running PBREAKL has shown that the rupture rate is most sensitive to thebreak diameter and the bubble point pressure. When carrying out simulations it is recom-mended that various sensitivity runs are performed in order to identify the range of results andto identify critical parameters. The amount of non-condensible gases in the composition, suchas nitrogen, should be checked as this can have a large effect on the bubble point.

PBREAKL is a relatively easy code to use but does have limits of applicability which are listedbelow. Violating these assumptions may not necessarily lead to poor results, but indicates thatthe code is being operated outside the range for which it is designed. If the assumptionsinherent in PBREAKL are too restrictive a more sophisticated code such as PLAC or OLGAshould be considered. This is particularly true if detailed topographic and thermodynamiceffects are important.

9.2.1 Limits of applicability of PBRJZAKL

l The pipe must be long compared to its diameter (length to diameter ratios greater than 100are recommended).

l The flow conditions upstream of the rupture are largely unaffected by conditions downstream of the rupture.

l The flows can be represented by homogeneous lumped masses.

l Pressure wave and flow inertia effects can be neglected.

l Each pipe section can be considered straight with no significant discrete loss components inthe pipeline.

l Friction losses at the rupture can be neglected.

l Conditions in the two-phase region of flow are governed by an isenthalpic flash process.

l The inlet boundaries can be represented by a constant mass flow.

l The flow through small breaks are unlimited by compressibility effects at the rupture.

l The rupture does not vary in size during the analysis.

l Heat exchange to or from the pipeline can be neglected, isothermal flow is assumed.

l The contents of the pipeline are initially all above the bubble point.

l Some of the functionality of PBREAKL is presently available in the PC based loss andconsequence model CIRRUS which is supported by the HSE group.


-


9.3 Factors Affecting the Transient Analysis ofPipeline Ruptures

The transient analysis of pipeline ruptures and depressuring can become very complexdepending on what is required from the analysis and to what accuracy. In some cases simplecalculations may suffice, but in others large errors can result from the most minor assumptions.The analysis is usually required to determine the liquid and gas flowrates from a rupture withtime for the assessment of jet fires and total volume spilled, or for gas release dispersion calcu-lations and determining environmental impacts. The latter is usually in order to study the effec-tiveness of safety valves. When the fluids are mainly gassy the temperature drop during depres-suring can also be a critical aspect of the analysis due to the temperature limits of pipelinematerials and components.

Figure 9. Hilly terrain pipeline schematic

In order to illustrate some of the factors affecting the spill size from a rupture in an oil pipeline,consider the hilly terrain pipeline in Figure 9. A simple approach would be to assume that duringdepressuring the bubble point pressure is reached at the high points A and C and hence the oilspill size is equivalent to the pipeline inventory between these two points. This volume could bemodified by performing a flash calculation from the initial pipeline operating pressure to atmos-pheric pressure in order to account for the evolution of gas. In practice some liquid could besiphoned over the hills towards the break and on reaching the bubble point the level swell or“champagne” effect produced by gas evolution could result in significant quantities of extraliquid flowing into the dip ABC. The Champagne effect can be halted if downhill elevations arepresent, since the hydrostatic pressure increases and can rise above the bubble point. The


f


bubble point at the interface may be reduced by the previous evolution of the solution gas. It ishence seen that the analysis is becoming quite complicated without consideration of thermaleffects, flowing conditions or the characteristics of the rupture.

Most ruptures are assumed to be circular holes with orifice characteristics, where as in practicethe rupture may be jagged and the flow path may be obstructed by the surrounding soil, forexample. Because of the difficulty of analysing these effects a free jet issuing from a circularhole is usually assumed. The effect of the back pressure on the efflux rate can be modelled bythe codes but the fluid/fluid interaction between the jet and the surrounding medium is ignored.In most cases the critical two-phase flow models are of acceptable accuracy for the calcula-tions required, with the Morris method probably most accurate although not as widely used asthe Henry-Fauske method. It is often useful to investigate the upper and lower limits of the crit-ical flowrate using the homogeneous equilibrium and homogeneous frozen models respectivelywhich are available in PLAC.

Work is underway with some of the sophisticated transient two-phase flow codes to enable theindividual components of the fluid stream to be tracked. In the case of pipeline blowdown, thelighter components may be expected to leave the pipe in the gas phase during the early stagesof depressurisation, with the composition becoming progressively heavier with time. Anothersituation where composition effects may occur is also when liquid condenses in a dip, leavingthe less heavy components in the gas phase. Present developments with some of the codesare attempting to account for this non-equilibrium between the phases.



9.4 Rupture Case Study - Bruce Gas Export PipelineTemperature Effects Resulting from a Pipeline Rupture

The Bruce gas export pipeline is 32” diameter and 5.8 km long and transports processed Brucegas from the PUQ platform to a tie-in to the Frigg system, which consists of 360 km of 32”pipeline. In the event of a rupture of the gas export pipeline, an SSIV located 193 m from theplatform will close, protecting the facilities if the rupture is on or near the platform. The expan-sion of the gas at the rupture causes a drop in the temperature, and hence there was a require-ment to determine the temperatures at the valve when actuated, to check that the valve stemmaterial remained above the brittle temperature. It was also required to provide valve tempera-tures for the estimation of the starting torque on the spindle. If the valve stem has cooled suffi-ciently to become brittle, then when operated it may break, making it impossible to close thevalve and prevent the pipeline depressuring.

Under normal steady flowing conditions with a gas export temperature of 55 “C the tempera-ture at the location of the SSIV is 48.5 “C at a pressure of 151 bara. If a pipeline rupture occursthe gas will be expanded to a pressure of 12.0 bara, equivalent to the hydrosystic head ofwater. An isentropic expansion over this pressure drop yields a final temperature of -76 “Ccompared to an isenthalpic final temperature of -15 “C. The minimum allowed temperature ofthe ball valve stem material is -30 “C, hence it is seen that based on isentropic assumptions thegas temperature could be much lower than the minimum allowed. As the pressure drops liquidis condensed in the pipeline giving a liquid mass fraction of around 10% at the break condition.A depressuring line is shown on the Bruce export gas phase envelope in Figure 10, indicatingthat for most of the blowdown the fluid state is two-phase. This prediction is based on theOLGA simulation described below and includes friction losses as well as heat transfer.

The actual temperature of the valve components can be significantly higher than the flowingfluid temperature due to the heat transfer between the fluid and the inside surface of the valve,and due to the thermal inertia of the valve material which will require sometime to cool byconduction. In addition the surrounding sea water may play a significant role in maintainingtemperatures above the brittle temperature.

A dynamic two-phase blowdown simulation was carried out using the OLGA code to assessthe temperature of the inner wall of the valve. Initially the simulation was run to steady state todetermine the initial pressure and temperature along the pipeline and specifically at the rupturepoint and the SSIV. The rupture was assumed to be 15 m downstream of the SSIV. The steadystate OLGA predictions are shown in Figure 11 where the platform is approximately 6 Km fromthe Frigg tie-in.

In the first dynamic simulation the program was re-started with the pipe section downstream ofthe rupture location removed and a fixed pressure boundary of 12.9 bara imposed, hencesimulating a full bore rupture. In practice 1 second is allowed for the pressure to fall from thesteady line pressure to the seabed pressure, hence the rupture is developed over this time. Theball valve begins to close after 91 seconds and is fully closed after 121 seconds.

Figure 12 shows the temperatures in the vicinity of the SSIV where it is seen that a minimumfluid temperature of -32 “C is obtained during the blowdown and rises rapidly as the valvebegins to close. The temperature at the upstream face of the valve rises much more rapidlythan downstream as the hotter gas is supplied from the reversed flow in the pipeline. Theminimum inner valve wall temperature is -3 “C, which is significantly above the brittle limit, eventhough the gas temperature is at times below -30 “C (Figure 13 shows the corresponding pres-sure plots). Even if the valve fails to close Figure 14 shows that a long time would be required



for the inner surface of the valve to drop to -30 “C, and a substantially longer time would berequired for the valve stem to cool to this value due to conduction through the 6” thickness ofthe stainless steel ball. The predicted minimum temperature at the rupture point is -45 C usingthe OLGA code which is halfway between the value calculated assuming isentropic and isen-thalpic expansion processes.

A final point about the valve starting torque is worth mentioning. Valves are usually tested in apressure and temperature controlled facility in which the torque is measured with the wholevalve cooled to say -30 C at the pipeline operating pressure, in this case around 151 bara.There is a time lag between the initiation of the rupture and the start of the valve closure, bywhich time the internal pressure has significantly reduced when the valve is actuated, in thiscase down to 20 bara, hence the starting torque is much reduced. In addition, the main sealsaround the ball that contribute to the torque are usually shielded by a debris control ring whichwill provide some thermal lag. The bulk valve will also be significantly warmer due to the effectof conduction and heat transfer from the surrounding sea water.

100100

9090

8080

7070

3030

2020

10

0

/

1;

Depressuring line

6.96.94.2 4.2 % condensate by mass

-50-50 -40-40 -30-30 -20-20 -10-10 00 1010 2020 3030

Temperature (“C)

Figure 10. Depressuring line and phase envelope for Bruce gas

I6 Issue Issue 2 2 -- December December 19941994

BP Multiphase Des&a Manual Section 9. Pipeline Ruptures

152 7 Frigg UK Wye - Bruce PUQ: Steady StateTime 50001.9s

- 148 I I I I I I 10 1000 2 0 0 0 3 0 0 0 4 0 0 0 5 0 0 0 6 0 0 0 7 0 0 0

Distance (m)

34 0

e 30

Time 50001.9s

I I I I I I I

0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0 5 0 0 0 6 0 0 0 7 0 0 0

Distance (m)

Figure 11. Steady state OLGA predictions


ISection 9. Pipeline Ruptures BP Multiphase Design Manual

Fluid temperature at valve

60

o^ 4 0e

e20

g O

F - 2 0

-40

2;o

Time (s)

3;)O 4;)o

60

40

Fluid temperature upstream at SSIV

0 100 200 300 400

Time (s)

Wall temperature at SSIV

-10 1 I I I I0 100 200 300 400

Time (s)

Figure 12. Temperatures in the vicinity of the SSIV


BP Multiphase Des&n Manuali

Section 9. Pipeline Ruptures

160160hh 140 140

Pressure at valve

I200200

Time (s)Time (s)

160160

"0"0140140

T;T; 120 120&-&- 100 100&& 80 80zz 60 60kk 40 40

202000

Pressure upstream of valve

0 100100 200200 300300 400400

Time (s)Time (s)

Pressure downstream of valve160 160 --

"0"0 140-140-r;r; 120-120-&- &- IOO-IOO-&& 80-80-zz 60-60-kk 40-40-

20-20-00 II II II II

0 100100 200200

Time (s)Time (s)

300300 400400

Figure 13. Pressures in the vicinity of the SSIV



60

40

Fluid temperature

200

Time (s)

4o 1Wall temperature

-30 f I I 1 1100 100 200 300 400

Time (s)

Figure 14. Temperatures downstream of the SW with no valve action


BP Multibhase Desinn Manual Section 9. Pipeline Ruptures

References

(1) Henry R.E and Fauske H.K “The Two-Phase Critical Flow of One-Component Mixtures inNozzles, Orifices and Short Tubes”, Trans. ASME-Journal of heat transfer pp 170-l 87, may1971.

(2) Morris S.D and Daniels L.C “Two-Phase Flow through wellhead Choke Valves A State of the Artreview”, AERE report G.3710,Oct 1985.


Section 9. Pdpellne Ruptures BP Multiphase Design Manual


LEFT BLANK


Section 10. Duwnfloul

Section 10. Downflow

Design of single and two-phase gravitydownflow systems

10.1 Background

10.2 Basic Design Approaches

10.3 Single Phase Downflow

10.4 Annular Liquid Flow

10.5 Transition From Annular Liquid Flow

10.6 Operating Experience

Reference


Section 10. Downflow BP Multiphase Design Manual


LEFT BLANK


rBP Multiphase Design Manual Section 10. Downfloul

10.1 Background

Single and two-phase gravity downflow occurs in a number of areas in BP’s operations.The principal examples of such systems are:

l Produced water overflow lines from an offshore separator system.l Cooling water return lines from offshore exchangers and condensers.l Rain and firewater collection systems from, for example, offshore helipad decks.

Over the last 5 years the Multiphase Flow Group has been asked to investigate operating prob-lems and design methods for a variety of downflow systems. As the amount of design data forsuch systems was very limited, and their operation poorly understood, XTC sponsored aprogramme of experimental work at Sunbury.

A large scale transparent test rig was constructed and operated by the Multiphase Flow Groupto study downflow in a number of pipework configurations. This work is reported fully in (1).


Section 10. Dowr@ow BP Multiphase Design Manual


LEFT BLANK

-


BP Multibhase Design Manual Section 10. Downjlow

--

10.2 Basic Design Approaches

A number of components of the design of gravity flow systems must be specified.

l Liquid may be supplied to the top of the downcomer pipe by a header tank, or by a pipe ofthe same, or different, diameter.

l The bottom of the downcomer pipe may be open to atmosphere, or submergedbeneath liquid.

l The top of the downcomer pipe may, or may not, connect to atmosphere.

The behaviour, or flow regime, of the flowing liquid will depend on the liquid flowrate, thecombination of the above inlet/outlet options, and the diameter and routing of the actual down-comer pipe. The two stable regimes are single phase liquid, in which the whole of the down-comer is full of liquid, and annular liquid, in which the liquid flows as a film down the inside wallof the downcomer, with a central gas core.

The selection of the combination of inlet/outlet options will be influenced by the source of theliquid flow, the desired flow regime, and safety considerations related to pipework supportand/or failure consequence analysis.

There are five principal recommended design options, given the assumption of pipe dischargeto atmosphere:

l Single phase downflow from pipe supply.l Single phase downflow from header tank supply.l Annular liquid flow from pipe supply (top of downcomer closed).l Annular liquid flow from pipe supply (top of downcomer vented).l Annular liquid flow from header tank supply.

These options are illustrated in Figure 1 and discussed in detail in the following sections.

Single phase liquid No top vent

Pipe supply Header tank supply Pipe supply

Figure 1. Downflow design options.


With top vent

Header tank supply

5

Section 10. Downflow BP Multiphase Design Manual


LEFT BLANK



10.3 Single Phase Downflow

Downflow gravity systems can be designed to operate in single phase liquid flow. This type ofapproach is often applied to the design of high capacity firewater drainage systems. However,in most cases it is not possible to ensure full pipe flow at all times. For example, a firewaterdrainage system will generally be maintained in a dry condition. When the system is calledupon to dispose liquid it will go through a transient stage during which gas displacement willoccur.

There are two design cases that can result in single phase downflow. The difference betweenthe two is the source of the liquid. In both cases, once the gas initially present in the pipe hasbeen displaced, the flow is single phase liquid through the system.

10.3.1 Supply by Header Tank

If the liquid is supplied to the top of the downcomer by a header tank then it is necessary tomaintain a liquid depth in the tank high enough to produce the liquid flowrate required tokeep the downcomer inlet, and the whole of the downcomer length, flooded.

(a) For downcomer entry at the base of the header tank

To avoid gas entrainment, liquid height, h, needs to be:

h 2 0 . 8 9

where:

h = liquid height in vessel above outlet (m)D = outlet pipe diameter (m)G = mass flow (kg/s)pL = liquid density (kg/m3)

(b) For downcomer entry at the side of the header tank

G*h 2 0 . 8 1 1

(9 D4 P,‘)

where:

h = liquid height above the top of the downcomer entry.


Sectton 10. DownJlow BP Multiphase Design Manual

In order to determine the capacity of single phase downcomer systems supplied by headertank, an energy balance is applied between the inlet, point 1, and outlet, point 2.

P u,* P u,*P, + ~ + pgh = P, + ~ + 0 + friction Equation 1

2 2

The static pressures acting at points 1 and 2, P, and P,, are both equal to atmospheric pres-sure. U,, the downward velocity in the header tank, will generally be very much less than U,,the liquid velocity at the pipe outlet, and so the above equation reduces to:

P u,*pgh = ~ + friction

2

Friction losses are the sum of losses over the pipe, and over any fittings.

P u,* L P u,*Friction = K ~ + f - ~

2 D 2

f = Darcy-Weisback friction factorL = pipe lengthD = pipe diameterK = total number of loss coefficients for fittings.so:

P u,* P u,* L P u,*pgh = ~ + K - - + f--

2 2 D 2

gh=T[l+K+f;] Equation 2

-

Hence, it is necessary to ensure that the required capacity (i.e. outlet velocity U,) can beachieved with the available head, h.

As already discussed, a drainage system designed on the basis of single phase flow, willprobably pass through a stage where gas is being displaced by the liquid. The system mustthen be designed to handle slugging flow which will occur as gas is swept from the pipework(see below).


BP Multiphase Design Manual Section 10. Downjlow

10.3.2 Supply by Pipe

If the liquid is supplied to the top of the downcomer by a pipe then the design requirement issimply that the liquid flowrate is high enough through the downcomer to initiate flooding andprevent de-flooding.

On the basis of the BPX tests this flowrate is highly dependent on the geometry (routing) ofthe downcomer. For a straight vertical downcomer with a pipe inlet the required Froudenumber to initiate flooding is 1.07. For a downcomer that includes a horizontal section therequired flooding Froude number varied, with position of the horizontal section, from 1.21(section near bottom) down to 0.85 (horizontal section near the top of the downcomer).

The Froude number is the quantity:

where:

V,, = superficial liquid velocity (m/s)D = pipe diameter(m)

So in order to design for single phase flow it is useful to have a horizontal section of pipe inthe downcomer, near the top of the system. This will give the greatest turndown on flowratewhilst maintaining single phase flow.

During the period when the liquid flow is filling the downcomer there will be a transient stageof gas displacement from the line that may result in surging and slugging, and some mechanicalvibration to the pipe. This must be allowed for in siting pipe supports.

In this case, where feed to the downcomer is directly from a pipe, the full energy balance givenabove in Equation 1 still applies, but the static pressure at the top of the pipe, P, is not neces-sarily equal to atmospheric pressure.

If P, is greater than atmospheric pressure, then the capacity of the downcomer will beincreased to a value in excess of that given by Equation 2.

However, if the flowrate of liquid into such a system is reduced to a value below that given byEquation 2, then the pressure at the top of the downcomer, P,, will fall to a value below atmos-pheric pressure. As the system being considered here has no vacuum breaker (ie is notvented) there is the danger of a very low pressure developing in the downcomer.

This risk of developing low pressures is greater in a system where the exit is below sea level, asthere is no opportunity for air to be sucked in to the bottom of the downcomer to help breakthe vacuum.


Section 10. Dowrq%w BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Des&pa Manual Section 10. Downjlow

10.4 Annular Liquid Flow

Liquid flows down the downcomer as an annular film, with a gas core in the centre of the pipethat is open to atmosphere at the bottom of the downcomer. The downcomer pipe may, ormay not, also be vented to atmosphere at its top end. The liquid flowrates achievable aredifferent depending on whether there is a top vent.

Offshore water overflow systems, such as cooling water return lines, and produced waterdowncomers, are generally designed to operate in the annular flow condition.

10.4.1 Without Top Vent

This case occurs with a pipe supply to the top of the downcomer. The classical design methodused to ensure that a downcomer remains in annular flow is to maintain the superficial liquidvelocity (Vsl) below a critical value given by:

V,, < 0.35 I/ gd

The term V,, / I/ gd is referred to as the Froude number, Fr. So, for annular liquid flow systemsthe relationship becomes:

Fr c 0.35

The experimental work at Sunbury was aimed at determining the criteria for maintaining annularflow in various piping configurations, and for downcomers whose exits are above sea level, andfor those which are submerged.

The experimental studies illustrated a number of points that had not previously been quantified.These are:

l The capacity of a vertical downcomer is significantly greater than that of one which hasbends and dog-legs.

l The capacity of a downcomer is reduced by submerging the exit below the water line.

Note that the capacity of a downcomer in this context is defined as either, or both of, the liquidflowrate leading to flooding, i.e. liquid fills at least part of the downcomer, or the liquid flowratewhich causes air to be drawn down the downcomer.

Based on the work at Sunbury, our recommended maximum flowrates which avoid floodingand air removal are as follows:

Downcomerconfiguration

Maximum Froude No.exit above sea level

Maximum Froude No.exit submerged

Vertical 0.57 0.3

With bends/dog-legs 0.35 0.3

It can be seen that the presence of bends significantly reduces the downcomer capacity.Ideally, bends should be avoided. If bends are unavoidable any near horizontal section of


Section 10. Downfow BP Multiphase Design Manual

pipework should have a minimum inclination of 1 in 40. To avoid slugging in any inclined lines itis usual to specify that lines run no more than l/2 full of liquid (for pipe internal diameters < 0.2m). For pipes > 0.2 m in diameter, the line can be allowed to run up to 3/4 full. Liquid depthsare calculated using the relationship :

v, =

where:

V, = mean actual liquid velocity (m/s)g = 9.81 (m/s*)m = hydraulic mean depth = area for flow / wetted perimeter (m)i = inclination of pipe to horizontale = pipe roughnessk = kinematic viscosity (mz/s)

With this type of system in which only the bottom of the downcomer system is open to atmos-phere, the gas (air) in the core is only circulated very slowly.

10.4.2 With Top Vent

This case can occur with a vent line put into the top of the downcomer supplied by a pipe, andwith a header tank system in which the flow is insufficient to maintain a liquid level in the tank,or if the tank has a hollow annular flow stabiliser. The behaviours of the systems are, however,very similar.

The experimental work at Sunbury illustrated that allowing air to be drawn down the down-comer in a top-vented system increases the amount of liquid that could be handled, i.e.increases flowrate to a level in excess of the rates given in the table in Section 10.4.1. Air isdrawn down through the core by the liquid flow, thereby giving a rapid replacement of the core.

-For a header tank without a hollow annular flow stabiliser the flow limit will be reached when theliquid bridges the top of the downcomer in the tank.

For the cases with a vent line (vented pipe supply, or header tank with hollow annular flowstabiliser) the limit will be the onset of surging caused by some liquid bridging in the down-comer.

In the experimental work at Sunbury, as the liquid flowrate increased, a point was reachedwhere a slugging/surging type of flow was produced.


BP Ah&phase Design Manuali

Section 10. DownjZow

10.5 Transition From Annular Liquid Flow

As the liquid flowrate is increased in a system operating in annular liquid flow, a point will bereached at which liquid bridging occurs at one or more locations in the downcomer. Liquidslugs are formed that move downwards through the system.

10.51 In Non Top-Vented Systems

In systems that have no top vent the slugs will begin to pull a vacuum as they fall, beginning theprocess of gas displacement from the downcomer. The gas is sucked downwards, and itsprevious position occupied by the liquid flowing into the top of the downcomer pipe. Someslugs of liquid will propogate right through the downcomer, drawing gas with them out of thebottom of the system. Other slugs will break up, leaving a churn flow region in the downcomer.

If the liquid flowrate is increased sufficiently, a single phase liquid flow will eventually beproduced. The instability of annular flow at these higher liquid flowrates is amplified if there areany restrictions in the downcomer, or reductions in diameter.

In a practical situation this type of flow might also cause liquid in any of the lines feeding thedowncomer to surge intermittently (as the slugs decrease the pressure at the top of the down-comer) and this, in turn, could affect the operation of the plant producing the liquid. Apart fromthe effect on the plant feeding the downcomer, slugging in the downcomer will produce out-of-balance forces which will require the pipe to be well supported, especially at bends. If a down-comer is operated at rates above the maximum values given in 10.4.1, slugging flow can beexpected.

Calculation of the velocity of a falling slug requires solving a complex force balance. TheMultiphase Flow Group, BP Exploration should be consulted if such a calculation is required.Once a slug velocity is determined the force on any bend is calculated using the equation givenin Section 3.4.1.

10.5.2 In Top-Vented Systems

In systems with a vent into the top of the downcomer, the effect of liquid bridging on the plantsupplying the liquid is much less of an issue. The pressure at the top of the downcomerremains constant at atmospheric. The only change will be in the gas rate drawn into the down-comer through the vent.

Any liquid slugs formed will, however, still have the potential to cause vibration to the pipeworksystem, and the caution on design of supports, as given in the previous section, still applies


ISection 30. Dowrlfrow BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 10. Downjlow

10.6 Operating Experience

If there is any chance that hydrocarbon vapours could enter the system (for example, throughentrainment in the cooling water from a leaking exchanger) it is essential that the entire down-comer retains the central gas core so as to enable explosive gases to escape through a saferoute to atmosphere. In the design of the Bruce platform, the FEED contractor undersized thedowncomer at 48”. When it was realised that a 60” downcomer would be needed to maintaina self-venting system, it was too late to alter the design. As a consequence, hydrocarbonvapour detection equipment had to be installed in the downcomer to warn of the presence ofany hydrocarbon gases being drawn down through the system as a result of a cooling waterexchanger leak.

In the Clyde produced water system the liquid flowrate was in excess of the values given abovein 10.4.1. As a result slugging and surging conditions were experienced, and this led to inter-mittent backing up of liquid in the lines that feed the system.


ISection 10. Dowvlflow BP Multiphase Design Manual

-

-


LEFT BLANK

-


BP Multiphase Design Manual Section 10. Downj.low

Reference

(1) Single Phasenwo Phase Downflow Studies, Brister, D, and Turner, A L, Sunbury BranchReport 124 179 dated 28.6.91


Section 10. Downfloul BP Multiphase Design Manual


LEFT BLANK

-


BP Multiphase Design Manual Section Il. Piping Layout

Section 11. Piping LayoutMechanical Design of Pipelines and Supports forMultiphase Flow

11.1 Converging Networks

11.2 Diverging Networks

11.3 Forces on Bends

11.4 Diameter Changes, and Fittings

11.5 Elevation Changes

11.6 Slugcatcher/Separator Inlet Nozzles

11.7 Liquid Distribution Through Bends

11.8 External Fluid Flow

References


Section II. Piping Layout BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 11. Piping Layout

11 Piping Layout

This section describes a number of ways in which the pipework layout may influence the behav-iour of the flow, and other ways in which the flow will affect the behaviour of the pipelines them-selves.

The influences of pipeline inclination angle, and of line diameter, are described in Section 3, inthe text on flow regimes and slug flow. This section (11) contains the current understanding onmore geometrically complex issues.


Section I I. Piping Layout BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 2 1. Piping Layout

11.1 Converging Networks

Pressure loss calculations for converging networks may be made using the MULTIFLO pro-gramme.

The principal issue here is how the upstream flows from separate lines influence the behaviourof the combined downstream flow.

Each line section in the network should undergo a flow regime prediction based on the in-situgas and liquid superficial velocities in that section. Where the upstream flows are predicted tobe not intermittent, but the downstream flow is intermittent, then the lines into and out of thepoint of combination may be treated as separate.

However, if one or more of the upstream lines is predicted to be in slug flow then this may havean effect on the nature of the downstream flow. If the downstream flow is predicted to be strat-ified then it may be forced into slug flow because of the incoming slugs. This case may occurwhere the downstream line is of a larger diameter than the incoming lines. If the downstreamline is predicted to be in slug flow then the slug frequency in this line may be affected by thefrequency of the incoming slugs. This will be most likely in systems where the incoming slug-ging line is the same diameter as the outlet line.

The main example of converging networks of interest to BP is in the Kuparuk field in Alaska (1).During one period of observation, slugs arriving at a junction of two 16 inch lines with a 24 inchout flow line were seen to decay, not propogating down the 24 inch line.

Further work is required to quantify the areas of the flow regime map (for a line downstream of ajunction of two or more lines) that are susceptible to influence from upstream conditions.


ISection 3 1. Piping Layout BP Multiphase Design Manual


LEFT BLANK

-



11.2 Diverging Networks

11.2.1 Occurrences

In general the main requirement of flow division by a diverging network is to split the incominggas and liquid flow equally between two branches of the same diameter. This occurs whereflow from a pipeline or manifold needs to be divided between two or more separation trains,and also in the inlets of finger-type slugcatchers.

Another instance of diverging networks is found in gas transmission systems, in which smallamounts of condensate may be present - the requirement is to know where the condensate willappear. This second example is not of current interest to BP. The key research work on thistopic has been done by Oranje (2).

11.2.2 Equal Division of Flow

For a diverging network, the key points to note are that in order to divide a multiphase flowequally the following provisions must be made:

l A dead-end tee configuration must be used, with the two exit arms both having the samediameter

l The inlet line must be straight for at least 15 diameters upstream of the tee

l The exit arms should be horizontal, with subsequent downstream pipework also horizontal ifpossible

l Exit arms should follow parallel routes downstream, to maintain equal pressure loss

l If the exit arms require a significant (lo-15 D) downwards elevation change after the tee thena line diameter equal or larger than the inlet line to the tee should be used, to preventpotenttial imbalance due to bridging (see Section 10) in the downflowing lines

l Exit arms should not be routed upwards after the tee

These recommendations have arisen out of three pieces of work conducted in BP. Firstly, theevaluation of flow splitting problems on Forties platforms Alpha and Delta (3); secondly, theevaluation of a design for a new production manifold on Forties Alpha (4); thirdly, a generalstudy of splitting at a tee at varying distances downstream of a bend (5).

These tests involved a number of different tee/bend configurations, as illustrated in theisometric drawings below.


rSection I I. Piping Lqout BP Multiphase Design Manual

Figure 1.

If the exit arms rise soon after the tee it is possible that very slight imbalances in the liquid headin the rising sections may feed back to the tee and affect the splitting. If the arms drop after thetee it is possible that a siphon effect may be set up, also affecting the flow division at the tee. Itis therefore recommended that the exit arms be made horizontal, or downwards of larger diam-eter if a significant downward elevation change is required.

A bend in the same plane as, and close to, the tee may cause very significant maldistribution,especially of the liquid (up to 61). To eliminate this effect the inlet line to the tee should bestraight over a length of 15 diameters.

To ensure good operability over a wide range of gas and liquid flowrate combinations, all thesedesign recommendations should be followed.

8 l.s.sue 2 - December 1994

BP Multiphase Design Manual Section II. Piping Layout

11.3 Forces on Bends

It is necessary to know the forces that might be applied on pipework by internally, and exter-nally, flowing fluids, in order to be able to design adequate supports and restraints.

11.3.1 Internal fluid flow

Any fluid or moving object that is constrained to change direction requires a force to be appliedto make it change direction. If the fluid has a steady density, and is moving at a steady velocity,then the force required is constant, and may be evaluated from a force/momentum balance.

For multiphase flow pipelines in stratified, bubble or annular flow the fluids have a reasonablysteady density and velocity. The changing momentum of the liquid phase is the principalcontributor to the force exerted on bends. In these cases the liquid momentum is not thatgreat as the flowing velocities are not very high. In a bubbly flow pipeline it would be veryunusual for the liquid velocity to be above 3 m/s. In annular flow and some high flowrate strati-fied flows the fluid velocities are much higher than that of bubbly flow. The gas velocity is muchhigher than that of the liquid film. However, the average fluid density is much lower than inbubbly flow due to the small liquid fractions in such flows. A steady, and moderate, force isapplied by the pipe on the fluids to change their direction.

However, in slug flow there are two differences from the other flow regimes. Firstly, the flowdensity is not steady, but varies between a low liquid hold-up in the film region, to a high hold-up in the slug body. Secondly, the velocity of liquid in the slug is roughly equal to the combinedgas and liquid superficial velocities. Therefore the force required to change the direction of aslug (high velocity and density) is much higher than the forces required during the otherregimes. In addition, the intermittent nature of the flow means the force is coming on and off atfrequent intervals. This can lead to pipework vibration and the possibility of fatigue of thepipework supports.

The slug frequency and holdup may be predicted using the methods available in the BP multi-phase design code, MULTIFLO (shortly to be re-issued as the UNIX code THOR). The detail onthe calculation of forces due to slugging around bends is given below in Section 11.3.2.

There have been instances (Prudhoe Bay, Alaska) of slug flow causing pipelines to jump off theirsupports. These cases occurred at expansion loops (a series of four 90 degree bends in a hori-zontal plane). The force required to change the direction of the slugs is provided by a flexing ofthe bends, which are not fixed to their supports because of the thermal expansion requirements. In some cases the movement of the pipe because of the flexing bends has been sogreat that the pipe moved off its supports.

In most cases the cause of the pipe leaving the supports could be attributed to unusual opera-tions rather than normal slugging flow. In these cases a shut-in (but pressurised) line wasopened to a depressured vessel causing rapid acceleration of the liquids accumulated in thelow sections. The forces that can be generated are greater than a reasonable support designcan accommodate. The problem was solved using operational procedures as well as safetyinterlocks to prevent opening major valves with an unbalanced pressure.

In one flowline the piping jumped from the supports during normal operations on three separateoccasions. This line was a tie-line diverting excess production from one production facility toanother. The line was operating at high velocities and would have been predicted to be oper-


Section 11. Piping Layout BP Multiphase Design Manual

ating in annular flow. The velocities were higher than would have been considered for slugrestraint design. One possible theory is that a slug developed in one of the individual flow-linesfeeding this tie-line and was passed through the tie-line at these high velocities. This scenariohasn’t been proven but it highlights the need to consider high velocity slug flow in similar “network’ systems.

Pipe vibration is also noticeable on Forties Alpha where the 12 inch line from Forties Echopasses through a bend upstream of the pig receiver, and then on into the production manifold.

In addition to rigid near-horizontal lines, the other systems in which the forces exerted by slugsare significant are those involving flexible risers. Flow-induced vibrations may affect the life ofthe riser, and of the riser connections and supports. For reference BHR group have undertakena number of studies of severe slugging in flexible risers. Further discussion on the behaviour offlexible risers under forces exerted by internal (and external) fluid flow are to be presented inSection 20 of this manual (Multiphase Flow Through Deepwater Systems).

11.3.2 Forces due to slugging

(a) Near horizontal lines

The force, Fg, exerted by the flow of gas only through a bend would be given by:

F, = 1/2 p, V,’ A

where:

Fg = force exerted by gas only (N)p, = gas density (kg/m3)V, = gas velocity (m/s)A = pipe cross sectional area (m*)

Gas Flow -Vm

Figure 2

If the pipe bend is carrying liquid only flow then the equation to determine force exerted is of thesame form, but written as:

F, = 42 p, V,2 A

-


BP Multiphase Design Manual Section I I. Piping Layout

where:

F, = force exerted by liquid only (N)p, = liquid density (kg/m3)V, = liquid velocity (m/s)

with p, > p, and V, < V,, and a typical F, usually greater than a typical Fg.

When slug flow occurs in an essentially horizontal line the mean velocity of the liquid in the bodyof the slug is equal to the mixture velocity, Vr,,, hence:

where:

V, = mean slug velocity

It is this velocity, VS, which should be used when evaluating the forces imposed by a slug as ittravels through a bend.

As a slug passes through the bend at velocity V,, the total composite force exerted on thebend, F,, is given by :

where:

pS = slug density

with p, - p,, V, - Vg, and therefore F, > F,.

If the slug hold-up is H,,, then pS is given by :

f-G = pI H,, + P, (1 - “1s)

The difference between the two steady forces F, and F, is the difference between the forcerequired to change the direction of the fluids in the predominantly liquid slug, and the forcerequired during the gas bubble. This is a little conservative as the gas bubble will usuallycontain some liquid flow. On slug arrival into the bend this change in steady force required istermed the slug force, F,, and is given approximately by:

F, = d2 b, - P,) V,2 A

Depending on piping configuration a considerable portion of this force may result in bending thepipe, giving a ‘pipe-spring’ which will then be released when the force due to the slug passageis removed as the slug passes out of the bend.


Section II. Piping Lqout BP Multiphase Design Manual

Fg

Slug out of bend --tVS

FsPipe spring force

Figure 3

These forces are based on normal quasi steady-state slug flow. They are not adequate toanalyse the dynamic problems previously described at Prudhoe Bay. In the largely dynamicsituation mentioned there are also large unbalanced pressure forces that occur as the slug isaccelerated.

It should be noted that in determining design loads for fixed supports to withstand these forcesdue to slugging, a dynamic factor of 2 (minimum) should be used unless a rigorous dynamicanalysis is performed (as discussed below). This results in:

F, = 2 42 (p, - p,) Vm* A

The dynamic effect of slug loads on bends is illustrated in the following figure. This shows aload cell output for a bend subject to slug flow. Note the short duration dynamic load (roughlytwo times the steady load) when the front of the slug enters the bend, and also when the end ofthe slug leaves.

The nature of slug flow means that these forces produce a tendency for the bend to move firstlyin one direction, as the slug enters the bend, and then in the opposite direction as the slug exitsthe bend.

If the piping is not adequately designed then these forces may cause excessive pipe movementor stress, and this may make the pipe and supports susceptible to movement induced fatigue.

(b) Pipeline Risers

It was mentioned in Section 3.3 that although a slug will progress through a riser at the downstream end of a pipeline, it will tend to decelerate as the line upstream of the slug packs up toprovide the pressure to overcome the increasing hydrostatic head in the riser.



1000 -

8 0 0 -

- Density

- - - - - Bending stress

- 0 . 8

- 0 . 5

- 0.4

8 0

Time (seconds)

Figure 4. Slug flow around a 90” bend.

As the slug leaves the riser the hydrostatic head loss in the riser reduces so that the upstreamgas bubble expands and accelerates the slug into the process plant. This phenomenon isdescribed in more detail in Section 3.

In order to model the effect of slug flow in a pipeline- riser, BPX have developed a dynamicprogram known as NORMSL. This can be used to determine the velocity of the slugs as theypass through the riser and into the topsides pipework. This information can then be used in thedetermination of forces exerted.

In order to assess the interaction of slugging with the topsides process plant the output filefrom NORMSL can be used as the input to a dynamic simulation model of the topsides plant.Process dynamic simulations can be performed using such software as the SPEED-UP andPLOT programs, and there is now a link between NORMSL and both SPEED-UP and PLOT. Inthis way any pressure changes in the process plant arising from the production of slugs or gasbubbles are fed back directly into the pipeline-riser model, thereby giving a better prediction ofthe slug velocities. This feedback effect becomes more significant as the height of the riserincreases, i.e. for deep water developments.

(c) Areas for Further Work

Analysis of test rig data is continuing and will be used to improve the slug frequency model,which will affect fatigue predictions.

Collection of data from hilly terrain pipelines operating in normal slug flow is required to vali-date/modify the existing models for slug frequency and velocity. Collection of data from theCusiana field should go some way towards achieving this aim.

A very simple model is currently used for determining slug liquid hold-up. Work is required todevelop a model which more closely reproduces the observed data. In particular the model willneed to account for the water content of the oil.


Section Il. Piping Layout BP Multiphase Design Manual

11.3.3 Support and Restraint Design Notes

For the initial design there is often adequate data providing the piping engineer where to lookand what to look for. The biggest problem is to ensure that the piping engineer knows that theline will be dynamically loaded. In many cases’ field problems are the result of either breakdown in communication between process and piping engineers or unexpected dynamic loadsthat exceed or visibly test the inherent factors of safety. Hence it is important to ensure an eft-cient transfer of data and information between the fluid system designer and the pipingdesigner.

There are several analysis techniques available to avoid or analyse these dynamic loads, andsome dynamic effects may be approximated by static loads.

The piping layout should contain adequate guides and line stops where practicable and longunguided sections should be avoided.

The piping layout connecting to relatively rigid pipework and equipment (manifolds, vesselnozzles, etc.) should be carefully designed to protect these locations from the dynamic slugging

forces. Possible solutions include:

l Adequate rigid restraints and anchors to protect these locations.

l Ensuring that force points (bends, etc.) are some distance from the critical location allowingboth piping flexibility and less robust restraints to provide protection for the equipment.

Where reasonable, piping should be supported or restrained at all heavy masses such asvalves, etc.

Include a dynamic factor of 2 minimum on design loads for fixed supports when using simplestatic analysis to calculate forces.

Avoid or minimise number of spring hangers.

11.3.4 Piping Stress Design Notes

l Do not use un-reinforced branch connections or fittings.l Do not use threaded connections.l To perform a quasi-static analysis:

Calculate momentum slug forces as previously outlined. If conservative values for densityand velocity were not used in calculating the slug force an additional 25% safety factorshould be included. (Note that velocity is the most critical factor and as a result determininga safe but reasonable maximum velocity is very important).

Perform steady state stress analysis of the possible load cases. There are several differentforce combinations depending on slug length and location of the slug (arriving and departing).For a piping system with n bends there are n load combinations assuming a single slugprovided it is possible to have slug lengths longer than the distance from the first bendto the last bend. Several of the cases may be trivial or redundant so that a limited number ofthe cases require further analysis.


BP Multiphase Design Manualii

Section 11. Piping Layout

Total stresses should be calculated and compared to allowable stresses. The cyclic (slugforce) and steady state components of stress should be calculated for use in fatigue life calculations.

Calculate permissible fatigue cycle-life using the appropriate code (6s 5500 for the UK) andsteady state and cyclic stresses calculated as outlined above. Estimate fatigue life based onslug frequency and duration.

This analysis does not include the possibility of resonance. Resonance due to normal quasi-steady-state slug flow is unlikely in that the slug frequency is not constant even during steadystate flow. Slug lengths and frequencies are very distributional rather than constant.

11.3.5 Field Problem Solving

When confronted with a specific case of pipeline response to the dynamic loads, a first step isto assess the need for immediate action. In a few cases the situation may be sufficientlyserious to cause a shutdown for correction or measures to limit the undesirable effects. On theother hand it is frequently possible and entirely adequate to reduce pipe movements by providing temporary braces, supports, ties, snubbers, or similar restraints, so long as care is taken toavoid possible thermal expansion difficulties on shutdown or during temporary operationalupsets.

More permanent corrective measures include an assessment of the movement(shakes) characteristics by field measurements and designing a permanent corrective solution.

Accurate measurements of piping deflections and frequencies can be used to gauge theseverity of the problem and determine whether action is required. For low frequency (very flex-ible) systems adequate measurements can generally be taken by simple manual methods(Don’t trust your visual impressions without actual measurement against some nearbystationary object). High frequency rigid systems require more sophisticated measuring toolssuch as accelerometers.

These deflections and frequencies can be used in a steady state analysis to estimate maximumand cyclic stresses. Areas of high stress intensification should be carefully assessed when considering fatigue live. These include branch connections, nozzles, equipment connections, girthwelds, etc.

At this point a permanent solution can probably be determined. It is possible that a morerigorous analysis is required; involving dynamic stress analysis, strain gauge measurements,etc.

There are several, well established, state-of-the-art piping flexibility programs, such asADLPIPE, Autopipe, PSA5, CAESAR II, CAEPIPE, etc. which can be used to evaluate thedynamic responses of the system and can be used during field investigations.

There are numerous other considerations (e.g. type of pipe material, local stress raisers, type ofwelds, strength of supporting structure, etc.) which are considered during the field ‘troubleshooting’. It must be pointed out that any field solutions to an existing piping system sufferingfrom the effects of dynamic loads should be preferably left to an experienced piping engineeringspecialist.




LEFT BLANK


BP Multibhase Desim Manual Section II. Piping Layout

11.4 Diameter Changes, and Fittings

This section is concerned with the forces exerted on pipework by multiphase flow passingthrough diameter changes and fittings. The significance of the diameter change or fitting isrelated to the flow regime, as was the case in the previous section.

Slug flow is the principal source of problems, especially where there are flow restrictions ordiameter reductions. The problems are two-fold. Firstly, a force is required to accelerate theliquid into a restriction and to overcome the increased pressure gradient through it - this force isprovided by an increase of pressure of the upstream fluids, and has a reaction that is met bythe pipe wall and the pipe supports. The reaction force is intermittent, matching the flowbehaviour, thereby creating the possibility of pipeworWsupport fatigue.

The second problem arises from the compression of the upstream gas. As described abovethe upstream gas pressure has to increase in order to accelerate the liquid into the restrictionand overcome the increased pressure gradient through it. The gas is therefore packed to ahigher pressure and density. As the end of the slug enters the restriction the acceleration forceis no longer required. When the length of slug contained within the restriction or section ofreduced diameter pipe starts to decrease (e.g. as the front of the slug passes out of the end ofthe restriction) then the pressure drop across the slug begins to drop. However, due to thepacking of the gas there is now excess pressure available upstream. This excess pressureaccelerates the slug.

The severity of the consequence of this is dependent on the position of the restriction relative toother restrictions or diameter increases. For a restriction (e.g. a partially closed valve) locatedsome distance (several slug lengths) away from other influences then the overall effect of a slugpassing through is just a force on the pipework containing the restriction, and some variation inthe velocity of the slug.

However, if there is a diameter reduction a distance of the order of a slug length upstream ofthe separator, then the effects may be serious. The upstream gas pressure rises to force theslug into and through the restriction. As the slug passes into the separator the pressure dropacross it reduces - but the upstream gas pressure then exceeds that required to overcome thepressure drop, and is therefore available to accelerate the remaining length of slug. This accel-eration continues until the end of the slug passes into the vessel. The upstream gas pressure isstill likely to be higher than that in the separator, resulting in a rush of gas into the separator.

This process is occurring in Kuparuk field, Alaska, where, in order to save money on the initialinstallation, two lines were reduced from 24 inches to 20 and 18 inches, for distances of 150and 30 metres respectively, upstream of the separator (1). The cost saving was achieved byhaving smaller valves. However, the separator internals have had to be replaced on sevenoccasions due to the mechanical damage caused by slugs being accelerated into the vessel.The gas compressors are also continually coming in and out of recycle to cope with the gasflow surges.

Designers of multiphase flow systems should minimise the number and extent of restrictions(and elevation changes), especially over the last few hundred metres of line. On no accountshould the line diameter be reduced within several slug lengths of the separator.




LEFT BLANK



11.5 Elevation Change

The effect of an inclination change is similar to that of a diameter change, in that the pressuregradient of the flow is altered. The consequences are described in the section on terrain andgeometry dependent slugging (Section 3.4.2)




LEFT BLANK



11.6 Slugcatchedseparator Inlet Nozzles

This section is an extension of Section 11.4 (Diameter Changes and Fittings) in that it deals withthe final section of pipework leading to the nozzle of the receiving vessel.

The main recommendation is that account be taken in the design of pipe supports of the loadimposed by slug flow. The number of bends close upstream of the inlet nozzle should be keptto a minimum. Ideally the inlet line should pass the fluids straight into the nozzle.

In Kuparuk field there has been an incident of weld cracking at a vessel inlet nozzle. The con-figuration of the inlet pipework is shown in the following figure. There are two 90” bends closeupstream of the nozzle. Slug flow through these imposes a bending moment due to the bendin the vertical plane, and a torque due to the bend in the horizontal plane. There is some scopefor compensation of the bending moment by flexing of the vessel shell. However, there is noscope for compensation of the torque other than by stress in the nozzle weld. Eventually thisweld failed.

It should also be noted that such a configuration may also result in maldistribution into the two‘halves’ of the vessel if the distance of the arm marked Z-Z is less than l&15 diameters (seeSection 11.2.2).

The remedy to this case was to construct a pipe support to take the loading on the bend in thehorizontal plane, thereby preventing any further movement and stress on the weld

Direction of flowDirection of force dueto bend in vertical plane

Side View

Plan View

Direction of flow

Direction of force dueto bend in horizontal plane

Figure 5. Kuparuk CPF-3 Case


Section 11. Piping Layout BP Multiphase Design Manual

An alternative inlet configuration is shown in the Figure 6, where there are no bends immediatelyupstream of the inlet nozzle. However, this type of vessel (end inlet) will have different separa-tion characteristics to that shown in Figure 5 (central inlet), and so the balance of requirementsbetween separation and inlet piping needs to be considered.To design for the pipe support loadings on any bends in the upstream pipework system it willbe necessary to produce force-time diagrams for the pipework configuration based on thenature of the slug flow through the pipework. The Multiphase Flow group can provide the nec-essary slug characteristics information.

Direction of flow

Side View

Direction of flow

Side View

Figure 6. Alternative configuration



11.7 Liquid Distribution Through Bends

The main area where this topic has provoked interest is in the distribution and stability of corro-sion inhibitor on pipewalls. At Prudhoe Bay problems have been experienced with excessivecorrosion just downstream of bends (in the vertical plane) of 30, 45 and 90 degrees (6).

Investigative work is at an early stage and recommendations on the optimum bend configura-tion have yet to be frnalised. Initial experimental work has been carried out for a road crossingconfiguration, where the pipeline drops down to go under a road. In this instance there isusually a liquid build-up at the base of the riser at the downstream end of the road crossing.

- Flow

Figure 7.

This liquid build-up reduces the gas flow path area, causing acceleration of the gas to a high in-situ velocity. This velocity is sufficient to entrain liquid as droplets, and accelerate them towardsthe gas velocity. The droplets then impact on the inside of the bend at the top of the riser. Ifthe drops contain any sand this will add to their inhibitor-stripping, and erosive, potential.

Bend angles (turning up from horizontal) of 90,45,22 and 11 degrees have been tested. Therestriction of the gas flow path by liquid accumulation occurs at each angle, with some dropletentrainment resulting. However, the extent of droplet impingement at the top of the bend thatbrings the line back to horizontal is reduced as the bend angle is reduced (7).

Further discussion of corrosion in multiphase systems is given in Section 21 of this manual(Corrosion Prediction in Multiphase Flow).


Section II. Piping Layout BP Multiphase Design Manual

-


LEFT BLANK

-



11.8 External Fluid Flow

The two main cases to consider are (i) wind flow over pipe suspended between two supports,and (ii) water currents flowing over unsupported sections of pipeline laid on sea, lake or riverbed, or around rigid or flexible risers.

For work on these aspects you will be referred to other groups in the Research and Engineeringorganisation.

Within fields of BP interest in Alaska there have been problems with wind-induced vibration. AtKuparuk field there has been a study of the use of helical strakes to reduce pipeline vibration.More recently damped mass restraints have been used. These consist of weights attached tothe pipe with specially designed elastomer bands.

The subject of the interaction of forces due to intermittent internal flows with externally gener-ated forces has not been a major topic of interest to date. Some work has been proposed onthe study of the behaviour of flexible risers under slugging flow, in which external currents couldbe considered.


Section Il. Piping Layout BP Multiphase Design Manual


LEFT BLANK



References

(1) ‘Multiphase Flow Field Tests in Kuparuk Field, Alaska, November 7th to December 1 st, 1990’,Sunbury Report No 21 164 dated 35.91.

(2) ‘Condensate Behaviour in Gas Pipelines is Predictable’,Oil and Gas Journal, July 2, 1973, 39-44.

(3) ‘Two Phase Flow Physical Modelling - Flow Splitting at a Tee’,Sunbury Branch Report 122 210, dated 30.11.83.

(4) ‘Physical Modelling of Multiphase Flow Splitting Effects at a Dead End Tee for Forties ArtificialLift Project’,Sunbury Branch Report 123 534, dated 10.3.89.

‘The Splitting of Two Phase Flow Regimes at an Equal Tee, at Varying Distances Downstreamof a Bend’,Sunbury Branch Report 123 395, dated 28.9.88.

(6) ‘Simulation of Multiphase Flow Characteristics Through the Sag River Crossing (Prudhoe Bay)‘,Sunbury Branch Report 123 713, dated 9.11.89.

(7) ‘Prudhoe Bay Road Crossings’,All-in-l document issued by CJ Nelson, dated 25.1.93.


ISection I I. Ptptng Layout BP Multiphase Design Manual

.THIS PAGE ISINTENTIONALLY

LEFT BLANK



Section 12. Solids TransportBehaviour of Solids in Multiphase Systems

12.1 Introduction

12.2 Solid Particulate Settling Characteristics andFlow Regimes

12.3 Model for predicting the limit of a stationerydeposit in horizontal and slightly inclinedmultiphase pipelines

References



-


LEFT BLANK


BP Multiphase Design ManualIll

Section 12. Solids Transport

12.1 Introduction

The hydrocarbons produced from a reservoir are sometimes accompanied by small quantitiesof solids such as sand or fracturing materials. For example, the Forties field produces in theregion of 5 to 40 pounds of sand for every thousand barrels of oil produced. This sand normal-ly collects in the separators and is either removed by manual intervention during maintenenceperiods or flushed out using a jetting system. When the South East Forties development wasconsidering using seabed templates connected to the existing Forties Alpha platform by two 5km pipelines, there was concern that the sand produced might settle out in the pipeline causingpigs to become stuck. The removal of sand may be relatively simple using pigging, providedthat only small amounts are deposited. The removal of larger quantities may be difficult andtime consuming. To design such systems required knowledge on how the sand is transportedand when it will accumulate. This prompted experimental work to be undertaken at BHRA in1983 (Reference 1). Again in 1990 the Forties Foxtrot development highlighted the need for abetter understanding of solids transport in multiphase pipeline systems.

In 1989 the three Ravenspurn South platforms were experiencing proppant production from thewell fracturing operation that led to the design and installation of separation facilities to preventthe carry-over of solids into the wet gas transport pipeline to the Cleeton platform. This studyrequired the prediction of conditions to transport the solids steadily to the separators withoutcausing undue erosion in the pipework.

Following from these studies BP has sponsored a project at the University of Cambridge toinvestigate the flow of solids in two-phase flow pipelines. Data has been collected for sandtransport in a horizontal and inclined 40mm ID pipe. This data has been used by theMultiphase Flow Group in XFE to develop preliminary design methods for predicting solidstransport conditions in multiphase flowlines.

The results of this preliminary work are presented here as a guide to predicting the critical con-ditions required to prevent solids accumulating in multiphase oil and gas pipelines. This is netessary to prevent pigs from becomming stuck and to prevent possible corrosion under soliddeposits in pipelines. Because of the possible stabilisation of solid deposits by heavy hydrocar-bons, and inhibitors, and the potential for accelerated corrosion under deposits, it is recom-mended to operate multiphase flowlines above the settling velocity to avoid solid deposition,and below critical erosion velocities to limit material loss.

This work is limited to solids which are heavier than the carrying fluids, ie sand and proppant,and may not be applicable to other solid substances formed by chemical reaction such ashydrates, asphaltenes and waxes.


ESection 12. Solids Transport BP Multiphase Design Manual


LEFT BLANK



cSection 12. Solids Transport

12.2 Solid Particulate Settling Characteristics and FlowRegimes

Flow regimes for solids transport in liquid/solid and liquid/gas/solid systems are illustrated inFigure 12.1.

(a) Stationary Bed

(b) Moving Dune

(c) Moving Dune / Scouring

(e) Plug Flow

(f) Slug Flow

(g) Low Holdup Wavy Flow

- - a

(h) Annular Flow

Figure 1. Flow regimes for the transport of solids in multiphase pipelines

Liquid/SolidSystems

Liquid/Gas/Solid- Systems


Section 12. Soltak Transport BP Multiphase Design Manual

12.2.1 Liquid/solid systems:

(a) Stationary bed

At very low liquid flowing velocities a stable solid bed is formed with particles at the bottom andno grains move at all. With an increase in the velocity a stable bed height is reached where theparticles at the top are transported further downstream to increase the length of the bed. Theupper surface of the bed is flat at very low flowrates but becomes wavy as the flowrate increas-es. At higher liquid flowrates the height of the stationary bed decreases. An equilibrium bed isreached when the shear at the upper surface of the bed transports solids downstream at a rateequal to the solid inflow rate.

(b) Moving dunes

If the liquid flowrate is increased further the bed breaks up and the particles arrange themselvesinto moving dunes in which the grains on the upper surface of the dune are rolled along fromthe shallow back region to the deeper downstream front, where they fall into the shelteredregion at the front . The dune passes over these particles until they are once again on the topsurface. The motion of dunes is similar to sand dunes in the desert and to snow drifts. Smallerdunes move faster than larger ones and a given length of stationary deposit will break up into anumber of dunes, each with a characteristic length and velocity.

(c) scouring

As the flowrate is increased further the grains roll along the top of the dunes with sufficientmomentum that they escape from the sheltered downstream region and are swept away asindividual scouring grains. Dunes can still survive in this erosion environment by replenishmentfrom upstream particles.

(d) Dispersed

At high liquid flowrates the dunes are dispersed and the flow consists of solids particles bounc-ing about in the flow and dispersed more in the liquid phase. A strong concentration gradient isstill however usually observed.

12.2.2 Liquid/gas/solid systems

Since the solids are heavier than the carrying fluids they are usually transported along the bot-tom of the pipe when the concentration is low. For this reason the flow patterns observed insingle phase solid/liquid flow are similar to those seen in stratified solid/gas/liquid flow since theliquid occupies the lower part of the pipe and the flowing velocity is steady. This is not howeverthe case when the gas/liquid flow regime is plug or slug flow, as the depth of the film and thevelocities vary.

(a) Plug flow

In plug flow the gas bubbles can flow along the top of the pipe and have little effect on thesolids flow with the full range of regimes already mentioned possible. As the amount of gas is


BP Multiphase Design Manual Section 12. Solids Transport

increased the bubble depth increases and the fluctuating velocities affect the transport similarto that described in slug flow.

(b) Slug flow

In slug flow the transport of solids is complicated as the solid may settle during the passage ofthe film region and may be transported in the slug body. There can be a large diameter effectas the depth of the film varies and shields the bottom of the pipe from the turbulence of theslug. A bed can be formed if the soild is not transported by either the slug or film. In caseswhere the solid is transported in the slug only the motion is obviously intermittent. The frequen-cy between slugs may be a factor if bed compaction and stabilisation by other products is apossibility.

For slug flow in slightly uphill inclined pipes the solid may be transported backwards due to thereverse flow in the film region, and hence the overall motion of the sand depends on the effi-ciency of the forward transport by the slug and the reverse motion caused by the film region.

(c) Low holdup wavy flow

In wet gas pipelines the liquid can be transported as a thin film along the bottom of the pipe, inwhich case the solid concentration in the film can be high, and in the extreme may appear as awet solid bed. In this case little is known currently about the conditions required to remove thewet solids.

(d) Annular flow

In annular flow the solids may be transported in the liquid film and the gas core. Since thevelocities are high in annular flow the usual concern is whether the erosion rate is excessiverather than if the solids will be transported or not.

Several factors can significantly complicate the analysis of the conditions required to preventthe accumulation of solids in multiphase pipelines. These include:

l Three phase flow effects (gas, oil, and water flow)l Preferential wetting of the solids by another phase ie water wet solids removal by the oil

phasel Bed stabilisation by other products, ie waxl Effect of inhibitors


Section 12. Soltds Transport BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 12. SolUs Transport

12.3 Model for predicting the limit of a stationary deposit inhorizontal and slightly inclined multiphase pipelines

The conditions required to prevent the formation of stationary deposits in multiphase pipelinescan be estimated using a method developed by XFE in 1993 (Reference 2). The model is basedupon a series of equations derived by Thomas (Reference 3) for calculating the friction velocityat the limit of solid transport in a liquid/solid system. The friction velocity is related to the pres-sure gradient and has been extended by XFE to the case of transporting solids in multiphasesystems by estimating the flowing conditions that give rise to the same pressure gradient that isrequired to transport solids in the liquid/solid system. The model is hence called the minimumsolids transport pressure drop model. The Thomas equations are used to predict the flowingpressure gradient associated with the minimum transport condition in liquid/solid flow whereenough energy is passed to a solid particle to enable it to remain in the bulk of the fluid phaseand to be transported downstream. Using this pressure gradient, a locus of points can plottedon a two-phase flow pattern map for a constant pressure gradient equal to the pressure gradi-ent at the minimum transport condition.

In the XFE model the two phase pressure gradient is predicted using the method of Beggs andBrill by guessing values for the gas superficial velocities for a given liquid superficial velocity andcalculating the two-phase pressure gradient. Iteration is performed until the velocities produce apressure gradient equal to that for the minimum transport condition caluclated by the Thomasequations for the same liquid flowing velocity. The calculation is repeated for a range of liquidvelocities to yield a locus of velocities above which the pressure gradient should be sufficient totransport the solids along the pipeline.

12.3.1 Determination of pressure gradient at the minimum transportcondition

Thomas derived several equations for the minimum transport condition depending on whetherthe solids particle diameter is smaller or larger than the laminar sub-layer in the liquid, anddepending on the solids concentration. The first step in the analysis is hence to determine thesolids particle diameter and the thickness of the laminar sub-layer, however, since the thicknessof the laminar sub-layer depends on the Reynolds number, some iteration is required. We initial-ly assume that the particle diameter is greater than the thickness of the laminar sub-layer andcheck for this condition after the friction velocity has been calculated.

12.3.2 Particle diameter

It is important to use the correct particle size in the analysis as this affects the calculation of theparticle settling velocity and also determines which method is used, depending on whether theparticle is smaller or larger than the laminar sub-layer. For single sized particles this is no prob-lem. However in practice the solids produced with oil and gas usually contains a spread of par-ticle sizes. Figure 12.2 shows the particle size distribution for Forties sand and shows that thesize varies from around 1 mm to 45 micron. For most cases it is recommended to use the meanparticle diameter or d50 value (in this case 255 micron) for the determination of the minimumtransport criteria. However it is also recommended to investigate the sensitivity of the results tothe particle diameter used.


Section 12. Soltds Transport BP Multiphase Design Manual

-

255

Particle size (+m)

1000 5000

Figure 2. Particle size distribution of Forties sand

12.3.3 Thickness of laminar sub-layer

The thickness of the laminar sub-layer is simply related to the pipeline diameter and theReynolds number for the case of a smooth pipe with a Reynolds number ~10’. The relationshipis as follows:

6 = 62 D Re7j8

where:

Re = [1488DV,J/p6 = thickness of laminar sub-layer in ftD = pipeline diamater in ftV, = superficial liquid velocity of interest in ft/sp = density in Ib/ft3J.,L = dynamic viscosity in CP

12.3.4 Particle settling velocity

The particle settling velocity is the velocity at which particles will settle under gravity in the stag-nant fluid. This velocity is primarily determined by the relative magnitude of the gravity and theviscous drag froces acting on the particle. Three settling laws are required to cover the possiblerange of settling conditions from low Reynolds numbers (ie small particle diameter/high viscosityfluid) to settling with high Reynolds numbers (large particle diameter/low viscosity fluid).

The three relationships for the settling velocity are as follows:-



(a) Stoke’s law

For Re, < 2

Ws = [1488 gd* (p,-p,) 118~1

(b) The intermediate law

For 2 < Re, < 500

Ws = [3.54 go.” d’ .‘4 (P,-PJO.“l/ [P,“.2g IJo.43

(c) Newton’s law

For Re, > 500

Ws = 1.74 [gd (p,-p,) / pJ”.5

where:

the particle Reynolds number is given by:

Re, = (1488dwsp,)/F

Since the Reynolds number depends on the particle settling velocity the correct equation to useis found by calculating the settling velocity and Reynolds number by each equation and com-paring the Reynolds number with the applicable limits for each method. For particles ofbetween 50 and 1000 micron in oil the appropriate law is likely to be the Stoke’s or intermedi-ate laws.

The particle settling velocity can be used to estimate the flowing conditions required to trans-port solids in vertical pipes. For liquid/gas/solid flow it is required to consider in which phase thesolid particles are transported.

12.3.5 Friction velocity at minimum transport condition

When the particle diameter is larger than the laminar sub-layer then the friction velocity at depo-sition for the limiting condition of infinite dilution is correlated by:

u; = [0.204 w, (v/d) (v/D)-O.~ {(p,-pJ / pJ-“.23]o.714

where:

ws = particle settling velocity (ft/s)

%* = friction velocity at minimum transport condition for infinite dilution(ft/s)d = solids particle diameter (17)v = kinematic viscosity (ft%)


rSection 12. Soli& Transport BP Multiphase Design Manual

When the solids concentration is high the friction velocity is modified by the following relation-ship:

(u,*/u,*) = 1 + 2.8 (ws /u,*)O.~~. @“.5

Where u,* is the friction velocity at the minimum transport condition for a given concentrationand @ is the solids concentration volume fraction in ft3/ft3.

For most cases of interest in oil and gas pipelines the solids concentration is low and the firstequation is usually sufficient to determine the friction velocity for minimum transport. Howeverthe concentration correction may be required if the liquid holdup is very small.

When the solids particle diameter is smaller than the laminar sublayer the expression for the fric-tion velocity at the minimum transport condition is:

U* = [l OOw, (v/d)2.71]o.26g

Given the friction velocity the Reynolds number and the thickness of the laminar sub-layer canbe calculated and the appropriate friction velocity expression checked.

12.3.5 Pressure gradient at minimum transport condition

Following the above procedure determines the friction velocity at the minimum transport condi-tion for the liquid phase. This is easily used to calculate the associated single phase pressuregradient at this condition using the expression:

Tlltc = (4 p, u’~) / [144 g, D]

where:

g, = 32.174

A two phase flow pressure drop method can now be used to determine the liquid and gasvelocitiy combinations which result in the same two-phase flow pressure gradient. It is useful toplot the locus of these points on a flow pattern map to indicate the conditions under whichsolids may or may not be transported. Alternatively comparing the two-phase pressure drop forthe conditions of interest with that for the minimum transport condition will indicate whethersolids are deposited or not. Figure 12.3 shows a comparison of the model predictions withsome of the data obtained at BHRA.

-

1

XFE have PC computer programs available to perform these calculations and to plot the results.For more information contact Paul Fairhurst on Sunbury 2983.


BP Multiphase Design Manual Section 12. So&Is Transport

0 Model - d -C delta, dP = 0.00389 psi/ft

z-‘cv ‘O-l w Experimental observations

c ,83; 0.1

$

g 0.012

0I I I I

0.10.1 11 1010 100100

Superficial gas velocity Vsg (ftk)

Chart 1 - BHRA Case A - 1” Pipe (air/water)

0 Model - d -Z delta, dP = 0.00180 psi/ft

2 H Experimental observationsaG lo--iii>> 0 l =ec 6% 0.) .

iil- ..I l m

P 00 n m

g O.l-3

. +=.Y 05 0c$ 0.01 I II II IIcncn 0.010.01 0.10.1 11 1010 100100


Chart 2 - BHRA Case B - 2” Pipe (air/water)

0 Model - d < delta, dP = 0.00668 psi/ft

G-n Experimental observations

g ‘O-

75>

. . 0,. 0.0l .>

n *.z l- 000%

0:

z0n m

3O.l-

p0) 0 05 0.01 I II I I IIrJYrJY 0.010.01 0.10.1 11 10 10010 100


Chart 3 - BHRA Case C - 2” Pipe (water/glycerol mixture)

Figure 12.3 Comparrisons of the model with BHRA data

Issue2Issue2 -- December1994 December199413

Section 12. Solids Transport BP Multiphase Design Manual


LEFT BLANK

-



(1)

(2)

(3)

References

Fairhurst,CP “Sand Transport in the South East Forties Pipe Line”, BHRA,I983.

Smith,M “A Model for Predicting Solids Transport in Near Horizontal Multi-phase Oil and GasPipe Lines”, XFE report 8/2/l 993.

Wasp, Kenny & Gandhi “Solid-Liquid Flow Slurry Pipe Line Transportation”, Gulf PublishingCompany, Clausthal, Germany 1979.


Section 12. Solids Transport BP Multiphase Design Manual

-

.THIS PAGE ISINTENTIONALLY

LEFT BLANK


‘-


Section 13. Three Phase Flow

Effect of Water on Oil/Gas andGas/Condensate Systems


BP Multiphase Design Manual Section 3 4. Flow Monitorha

Section 14. Flow Monitoring

Monitoring of Multiphase Flow

14.1 Instrumentation Requirements

14.2 Techniques AvaiIable


Section 14. Flow Monttortnn BP Multiphase Design Manual


LEFT BLANK

-


BP Multiphase Design Manual Section 14. Flow Monitorlng

Flow Monitoring

This section of the guidelines will first describe the requirements for instrumentation of multi-phase flowlines. The context of this is of data acquisition above and beyond the normalmeasurements of pressures, temperatures, and single phase flowrates after gas/oil/water sepa-ration. In these descriptions of requirements the section numbers of appropriate measurementtechniques are given for easy reference.

In the second half of this section brief descriptions of available techniques are given. Some arenon-intrusive, and may be fitted temporarily to systems. Other measurement techniques willrequire a spoolpiece or tapping in the line.

For monitoring of phase flowrates in multiphase flow, please refer to Section 15 (Metering).


Section I4. Flow Monitoring BP Multitdase Design Manual


LEFT BLANK

-



Section 14. Flow Monitoring

14.1 Instrumentation Requirements

14.1.1 Flow Regime Identification

The original requirement for flow regime identification in oilfield multiphase flowlines arose fromthe R&D activities set up to develop and verify flow regime prediction techniques. The design ofprocessing facilities is affected by flow regime, in that slug flow may cause unacceptable gasand liquid flowrate variations in the plant. Therefore reliable prediction of flow regime over awide range of gas and liquid flowrates, pressures, temperatures, fluid compositions, and linetopographies are very important. See Sections 1 and 3 of this manual for further discussion ofthis issue.

A series of field trials has been conducted within BPX to collect data on flow regime. This activ-ity will continue on major new oilfield developments to ensure that the prediction methodsremain applicable in new combinations of conditions.

Please refer to Sections 14.2.1 - 14.2.4 and 14.2.6 - 14.2.7 (for information on the differentmeasurement techniques available for this activity.

14.1.2 Line Holdup Estimation

The next step beyond identification of flow regime is to be able to determine the hold-up ofliquid in the pipeline. This is required in the assessment of pigged liquid volumes, and in theprediction of slug volumes generated by flowrate or topography changes. Some pressure lossprediction methods also rely on predictions of liquid hold-up.

As with Flow Regime Identification this activity is largely R&D based, to verify hold-up predictionmethods for future design requirements, rather than for a current or future normal oilfield opera-tion. For this activity the techniques described in Sections 14.2.3 - 14.2.7 may be appropriate.

14.1.3 Slug Flow Characterisation

This activity has two main thrusts - firstly the provision of data and subsequent verification ofslug flow design methods, and secondly the provision of information for feed-forward control ofslug reception facilities (slugcatchers, separators).

(a) Slug Flow Design Methods

The instrumentation requirements are threefold - to measure the frequency of slugging, thevelocity of slugs, and the phase volumes (gas, oil and water) contained within the slugs andsubsequently delivered to the process. Data acquired is used to validate the predictive tools forthese parameters.

For the development of slug flow design techniques it has been important to gather data over awide variety of conditions of gas and liquid flowrate, line diameter and topography, and fluidproperties. Continuing data acquisition on lines with new combinations of conditions will becarried out as appropriate. See Section 3 of this manual for an in-depth discussion of thedevelopment of the actual design methods. The current thrust is in the development and vali-dation of methods for hilly terrain systems.


Section 14. Flow Monitoring BP Multiphase Design Manual

A pair of measurement positions is required for the determination of slug velocity. Techniquessuitable for slug flow characterisation are described in Sections 14.2.1-14.2.4 and 14.2.6-14.2.7. The techniques in Sections 14.2.1-14.2.2 do not give the phase fractions within theslug.

(b) Feed-Forward Slug Flow Control

The requirements for this application are twofold - firstly to measure the velocity of slugs, andsecondly to determine phase volumes contained within the slugs. This allows determination ofthe arrival time of slugs into the reception facilities, and gives the liquid inflow rate value to anassessment of the likely effect of the slug on oil and water levels, and of the effect of any gassurge following the slug on the gas plant.

Appropriate instrumentation techniques are the same as those listed in the Slug FlowCharacterisation section.

Location of the measurement position has to be carefully selected in order to give sufficient timefor any feed forward activity (e.g. increasing pump-out rates) to take place. The biggest benefitwill be obtained from such feed forward control in systems where the slug frequency is quitelow, where a slug arrival will cause a significant change in separator vessel conditions. This willbe the case in larger diameter lines, and lines operating close to the slug-stratified boundary.

14.1.4 Reservoir Management

A knowledge of the volumes of fluids produced from different wells in a field is generallyrequired in order to be able to manage depletion of the reservoir. Water breakthrough needs tobe identified, as does gas coning. The current industry standard is to flow wells one at a timeto a test separator via a dedicated test line. The flowrates of gas, oil and water are measuredafter separation. Wells are tested typically once a month. Required accuracy is of the order of+/- 5 to 10% on each phase.

In an increasing number of satellite tiebacks to existing installations, the requirement to have alocal test separator, or a second pipeline back to a test separator at the host facilities, can addan appreciable cost to the basic development. The cost of metering for reservoir managementcould be reduced by using a multiphase metering system that is more compact, and cheaper,than the test separator system. Details of possible systems are contained in Section 15,Multiphase Meter Systems.

14.1.5 Production Allocation

In some developments with multiple fields the cost of installation of separate processing trainsfor each field may be prohibitively expensive. Subject to the agreement of all partners andregulatory authorities, it may be acceptable to allocate production back to the individual fields(to determine royalties and taxes) using methods that do not require full processing trains withfiscal metering.

In some cases a remote test separator will be used (e.g. Point McIntyre to Niakuk), with adjust-ments made to ensure that the sum of well allocations from all contributing fields equals thefiscally metered export rates from the central processing facility.


BP Multiphase Design Manual Section 14. Fbw Monitoring

In other cases a further cost saving may be obtained by using multiphase metering systemseither remotely, or just upstream of where the production from one field is co-mingled with thatfrom other fields (e.g. under consideration for RAP). The multiphase metering methodscurrently available for this activity are described in Section 15, Multiphase Meter Systems.

Current regulatory guidelines centre on the need for each contributing field to have the samemetering set-up, so that any systematic errors tend to offset for the different fields.

14.1.6 Fiscal Measurement

A further, even more difficult requirement, is to measure the phase flowrates in a pipeline suff-ciently accurately to allow for custody transfer to the same level of accuracy as current singlephase standards.

As yet no multiphase measurement technique or product is sufficiently accurate to achieve thisto single phase flow standards. Such accuracy requirements are of the order of better than +/-1%.

In parallel to developments aimed at improving the accuracy of multiphase metering systemswill be negotiations as to what is an acceptable accuracy for partners and governments associ-ated with an oilfield development. These two aspects (technology advance, and relaxation ofaccuracy) will probably both have to progress in order for ‘fiscal’ metering of multiphase flow tobecome a reality. Some precedent is already found on the North Slope of Alaska, where smallfields have their production allocated on the basis of well test results (as described in Section14.25). In this case there is no bulk flow measurement upstream of the point of co-minglingwith fluids from another field with different (slightly) ownership.


Section 14. Flow Monitoring BP Multiphase Design Manual


LEFT BLANK


-


iSection 14. Fibw Monltorttzg

14.2 Techniqu& Available

14.2.1 Pressure

The analysis of a pressure measurement made by a relatively fast pressure transducer mayyield some information. Differential pressure across two transducers has even more value.

It is possible to obtain an indication of flow regime from a pressure measurement. Simpleobservation shows the passage of slugs past a measurement location as a significant andpotentially prolonged rise in pressure. Pattern recognition can be used to analyse the signalfurther to differentiate between stratified smooth, stratified wavy and annular flow. Such ananalysis may be supplemented by flowrate or pressure loss information, which will significantlyreduce the uncertainty in the deduction. However, it is not possible to get information on liquidcontent in a cross-section from pressure only.

Forties Alpha riser base

Forties Alpha riser top

” I0 h

I4 A ; l b

Time (minutes)

Figure 1. Pressure fluctuations in Forties Echo-Alpha flowline.

Some extension of the use of pressure measurement for flow metering has been proposed.Firstly, the variation of pressure drop across a choke can give an indication of flowrate changes.This information can be used effectively during the periods between well tests to give approxi-mate flowrates for choke settings different from those under which well tests were conducted.Pressure loss down a long length of flowline also gives an indication of flowrate.

Secondly, some development of neural network systems to analyse data and ‘train’ a computerto interpret pressure signals as a flowrate measurement has gone through preliminary stages.The ability to make calibration (training) runs is required here.

14.2.2 Acoustic Monitoring

There are two basics approaches to acoustic monitoring - passive and active. In a passivesystem the measurement device listens to the noise generated by the flow. An active systemputs sound energy into the fluid - the sound is then picked up at a receiver after passage into orthrough the fluid.


Section 14. Fhu! Monitoring BP Multiphase Design Manual

(a) Passive Systems

One such system has been developed within BP Exploration. Stemming from the requirementsto have more easily portable and non-radiation based techniques, the idea of listening to theflow was pursued. The result is a hardware and software package that has been usedsuccessfully for flow regime identification and slug flow characterisation in a number of fields(Wytch Farm, Kuparuk, Prudhoe Bay...).

Transducers pick up flow generated noise around the 70 kHz frequency. After filtering thesignals can be analysed to give wave and slug frequencies and velocities, and slug lengths.

Stratified smooth flow regime Stratified wavy flow regime

Time Time

Slug flow regime

EbTime

Figure 2. Title ?

(b) Active Systems

Annular flow regimeJ

Time

These have been used for two different aims - flowrate measurement, and gas fraction determi-nation. Doppler frequency shift techniques are used for single phase liquid flowrate measure-ment. The principle has been tried in multiphase flow but with little success, because thegas/liquid interfaces break up the acoustic signal.

The fact that this break-up occurs has been used in a device to measure gas fraction near thepipe wall. By directing an acoustic signal into the flow, and measuring how much of the signalis scattered back, an estimate of gas fraction can be made (after calibration).


BP Multibhase Design Manual Section 14. Flow Mondtortng

14.2.3 Gamma Ray Densitometry

This is the most widely used instrumentation technique for studying multiphase flow character-istics. The principle is to direct a beam of gamma rays from a sealed source (usually Caesium137, although some dual energy systems have also been tried e.g. Caesium and Americium)across a pipe and to measure the intensity of radiation transmitted through the pipe wall, anyinsulation, and the fluids within the pipe. The absorption by the pipe wall and insulation can becalibrated out to give as output a measurement of the density of the fluids within the pipe.The technique will give a plot of density vs. time at the measurement location. From a tracetaken from a gauge mounted with the gamma beam vertically through the pipe, the flow regimecan be identified as slug, bubble, stratified smooth or wavy/annular. To resolve betweenannular and stratified wavy the gauge is then turned through 90” to transmit through the pipehorizontally. The existence or otherwise of a liquid film at the mid pipe location will definewhether the flow is wavy or annular.

Setting two gauges on the pipe separated by a distance allows cross-correlation of the signalsto determine the velocity of the slugs and waves between the measurement locations. Coupledwith the measured time taken for a slug to pass a gauge, this will yield a slug length.

T- 8oo 11

~JJ 600 -

s 4oo -2 200 -

6 0iI

I!i

Ilb

I15

Time (minutes)

-“E 800

3 6002 400.fe:2 200

8 015 20 25 30

Time (minutes)

Figure 3. O-l 5 minutes period and slug flow, 15-30 minutes period and wavy flow.

14.2.4 X-ray

The expense and size of X-ray systems has, to-date, prevented their widespread use in theoilfield environment. However, there have been isolated tests in which an X-ray system hasbeen used to identify flow pattern. Real-time displays are possible of the fluids in a pipe.

The principle of operation is the same as the gamma ray densitometer, in that the fluids in thepipe (gas, oil and water) absorb X-rays to different degrees. By directing an X-ray beam across


Section I4. Flow Monitoring BP Multiphase Design Manual

14.2.5

14.2.6

14.2.7

the pipe, and mounting a detector on the opposite side of the pipe to the source, it is possibleto ‘see’ the flow behaviour in the pipe.

Neutron Back Scatter

This technique depends on the concentration of hydrogen in the vicinity of the probe. Aneutron generating source is used to send neutrons into the process pipe/vessel. The intensityof emitted gamma rays from the interaction between neutron and proton in the hydrogen givesthe hydrogen concentration, which in turn indicates the presence of gas or liquid. This tech-nique can give an indication of flow regime, and also of liquid film depth in a pipe.

Capacitance/Gamma

The principle behind this technique is the determination of dielectric constant between elec-trodes by measuring capacitance, coupled with a gamma ray transmission measurement todetermine density. The measurement of capacitance requires great care in shielding anddesign of the electronics. The make-up of the overall dielectric constant is determined bymodelling the contributions of gas, oil and water in the three phase flow.

The output should give a fast response indication of the gas-oil-water phase fractions in thecross-section of a pipe. This will allow flow regime determination, and slug flow characterisa-tion if two devices are used with a suitable distance between measurement sites.

The systems are designed as full-bore in-line spools with ceramic liners. Further details may beseen in section 15, Multiphase Meter Systems.

Microwave Absorption

The principle behind this technique is also the determination of dielectric constant betweenelectrodes, this time by measuring microwave absorption, also coupled with a gamma raytransmission measurement to determine density. The measurement of microwave absorptionrequires great care in design of the intrusive probes. To resolve the individual phase fractions,the measured dielectric constant is compared with the value predicted from modelling thecontributions of gas, oil and water in the three phase flow. The calculation is iterated aroundthe phase fractions until the calculation matches the measured data.

The output should give a fast response indication of the gas-oil-water phase fractions in thecross-section of a pipe. This will allow flow regime determination, and also of slug flow charac-teristics if two devices are used with a suitable distance between measurement sites.

The systems are designed with intrusive probes. Further details may be seen in Section 15,Multiphase Meter Systems.


BP Multiphase Design Manual Section 15. Meteting

Section 15. Metering

Multiphase Meter Systems and their Application

15.1 Introduction

15.2 Multiphase Meter Development Review

15.3 Summary

References

Lssue 2 - December 1994

Section 15. Metering BP Multiphase Design Manual


LEFT BLANK


EBP Multiphase Design Manual Section 15. Metering

15.1 Introduction

In general, oil industry effort to develop multiphase flow metering techniques has been inprogress since the mid 1980s with one or two exceptions, such as various approaches byTexaco, which commenced a few years earlier. As a result of this industry-wide effort, keyphysical measurement techniques investigated are now emerging as commercially availablefield test hardware.

By comparison, relatively little attention has been paid to the problem of calibrating multiphaseflow meters. Some of the development projects have resulted in the construction ofoil/water/gas flow loop facilities for test purposes but little thought seems to have been given tothe problem of calibration of multiphase flow meters for field service.

This section of the Multiphase Design manual presents a review of multiphase meteringsystems in which BP is involved or the author has technical knowledge. This encompasseswhat are believed to be the most advanced and soundly based technologies. The review isnecessary in order to give a background perspective to the preceding statements concerningcalibration. It is also intended that it provide the state of the art of emerging technologies. Thereview includes information on current capabilities and costs as well as describing the differentmeasurement concepts.

The following section of the manual, Section 16, concerns calibration philosophy of multiphaseflow meters. The purpose of this is to highlight the key aspects and challenges faced in estab-lishing a calibration methodology for multiphase flow meters. Thoughts on the problem arepresented, particularly, with reference to calibration for single phase flow. It is intended that thecontents act as stimulus to trigger input to the problem in order to facilitate the development ofpractical calibration solutions.




LEFT BLANK

-


BP Multiphase Design Manual Section 15. Metering

15.2 Multiphase Meter Development Review

This review primarily covers technology to which BP has contributed directly and in which thecompany has had involvement. Comment is also included concerning other techniques ofwhich the author has knowledge and which have reached commercialisation and field teststage. It should be realised that the technology is currently evolving with new techniques, expe-rience and test results continually emerging.

Standard terms are used in quoting accuracy or error percentages. Specifically, relative errormeans error or deviation expressed as a percentage of actual flow rate. Absolute error refers toan error or deviation expressed as a percentage of full scale.

All of the metering approaches described below are targeted at accuracies for well test duty forreservoir management and the possibility of allocation of production between wells. The highaccuracy levels associated with custody transfer and the term “fiscal metering” in single phaseflow metering have not been addressed. On the basis of research and development experienceto date, this level of accuracy is not likely to be realised by multiphase metering techniques inthe foreseeable future.

15.2.1 BP-Patented Positive Displacement Screw Meter

Being licensed to ISA Controls Ltd. in 1994, this flow meter is believed to be the mostadvanced in terms of flow loop and field test experience. Two prototypes built to ANSI class1500 pressure rating and to NACE specifications in 1987 have been used for the testprogramme. A full range of flow conditions (of up to 95% gas void fraction (GVF) during trials atPrudhoe Bay) has been covered and several thousand run-hours experience gained. The metermeasures total volumetric and total mass throughput of the multiphase fluid stream and uses awater cut reading by some other means to measure the flow rates of oil water and gas.

Cost to supply a BP flow meter can only be estimated at present but is expected to be of order$150,000 for worst case service specifications. Cost of a complete ISA multiphase measure-ment skid will be application dependent (i.e. land or subsea packaging, number of meters togive turn-down, additional instrumentation required - see below). The two field prototypes weredesigned for a nominal pipe size of 3” or 4” and are approximately one metre in length andthree quarters of a tonne in weight. The opportunity to build the first commercial meter willpermit the incorporation of many design improvements realised through the course of theprototype testing since 1987. This could have significant implications for manufacture, ease ofmaintenance, size and cost.

Some limited study has been conducted addressing the costs of packaging and installing amultiphase metering system for subsea duty using the BP positive displacement meter. Thecastings are highly dependent on the application but demonstrate how the ex-works cost of themultiphase metering system is not necessarily the only significant contributor to the overallcosts associated with its implementation. The findings of these studies are available in two BPreports (1,2). Further development is required in order to package the sensors and signalprocessing and data transmission for remote service sub-sea, but this is not expected to be aproblem.

The BP meter requires an additional device to supply a reading of water cut in order to yield theflow rates of oil, water and gas. Currently, any one of three to four composition measurementtechniques under development could give the water cut measurement. The test work to date


Section 15. Metertn~ BP Multiphase Design Manual

indicates that these techniques can measure water cut by dynamically tracking the phase frac-tions in the pipe cross-section.

The test work has shown that the total volume positive displacement measurement can achievethe target of +/- 5% of reading (relative error). Currently, the corresponding turn-down variesfrom a minimum of 51 to 151, dependent on flow regime behaviour. The K-factor curve toenable this in multiphase flow can be determined on the basis of a calibration in a single phasewater flow loop.

The flow driven elements of the screw meter generate a pressure drop which depends on flowrate, gas volume fraction and flow regime. For high liquid loadings, pressure drops of up to 2.2bar have been observed during tests when the meter has been run at the top of its range.

The meter incorporates a commercially available gamma densitometer to measure mixturedensity. This may be factory calibrated on fluids of known density (such as air and tap water). Adensitometer of this type has been in use subsea, in order to monitor multiphase flow behaviourat the foot of the Forties Echo-Alpha 12” riser, since May 1988. This densitometer has requiredno subsequent intervention and is still functioning.

The principle behind the screw meter concept is that the positive displacement elementsconstrain the phases to move at a single velocity. The densitometer beam passes through acavity between the screws mid-way along the meter axis. In principle, if slip between thephases were fully eliminated between the screws, the densitometer reading would equate tothe true mixture density relating to the overall proportions of oil, water and gas flowing in thepipeline. Coupled with the water cut reading from a second instrument, the mixture density andtotal swept volume readings can then be used to solve out the flow rates of oil, water and gasat line conditions. (This assumes that the individual phase densities are known as a function ofpressure and temperature which are also measured at the multiphase metering section).

All the test data to date have shown a tendency for the mixture density (or total mass flowmeasurement of the stream) to exhibit a bias error which is a function of gas void fraction, GVF.Above 40% GVF, this bias depends on the installation site i.e. it is specific to the site. Currently,in order to achieve relative errors in actual flow rate of each phase of 5% to 10 %, this bias inmixture density or total mass flow rate must be calibrated out by reference to a measurement oftotal mass throughput at the site installation. This is not a regular calibration requirement butwould be necessary initially and subsequently following any change of operations which signiti-cantly influenced the nature (GVF) of the flow stream. Future development and design improve-ments are expected to reduce, and could eliminate, this requirement for a site calibration ontotal mass. (Note: 7/6/94: In 1993 the bias effect was eliminated in testing in Alaska).

-

Test data to date indicate that relative error uncertainty increases with GVF. With the biascorrection, the scatter of relative error in phase flow rate increases from 5% to 10% at amaximum GVF of 95%. If the mixture density or mass bias error is not site calibrated then themeter exhibits a general uncertainty in total mass flow rate increasing with GVF to +/- 15% at80% GVF. Without a total mass bias correction, the phase flow rate measurement errors canbecome large at high gas void fractions. The meter has not been field tested at the highestGVFs above 95%. Further experience would be required to confirm the GVF range of the meterat the highest levels. (Note 7/6/94: In 1993 testing at Prudhoe Bay with the bias effect elimi-nated, relative errors in phase flow rates were scattered f 5% . This was observed from manytests covering several wells and choke settings giving a GVF range of 75% to 95%).


BP Mu&phase Design Manual Section 15. Meteting

The above uncertainty levels relate to relative errors in total flow rate and individual phase flowrates. They have been determined on the basis of the scatter of measurement errors observedfrom all the test data (different sites and wells). It is possible that site specific calibration on totalmass flow rate on an individual well basis could exhibit a superior repeatability, thus reducingthe uncertainty from the above levels.

The bias error in mixture density is believed (on the basis of analysis of all the test data) to beattributable to flow pattern regime effects between the screws which do not, therefore, fullyhomogenise the flow and eliminate slip in the idealised manner. Further design modificationscould be investigated to address this which might eliminate the current need for a site specificcalibration. (See above).

The BP meter based system, including any of the separate devices being developed which willmeasure water cut, essentially, can be factory calibrated using single phase fluids in relativelystraightforward factory test cells.

Summary remarks:

The simple rugged mechanical design of the BP meter took full account of the need for manymonths service without maintenance in a flow stream exposing the meter to violent mechanicalshocks from slugging, abrasives and corrosive substances. The test programme so far hasconfirmed the design under such conditions. Experience will determine wear characteristics andthe extent to which these influence the need for re-calibration, if any.

A key reason for selecting the positive displacement design and adapting it to the no-slip prin-ciple was that this would result in a meter with relative insensitivity to the generally non-repro-ducible flow patterns of multiphase streams. It therefore follows that the principle should lenditself to calibration.

The current design requires an initial factory calibration. The question arises- is it desirable toprove the meter once in service? Is this even a practical proposition for a multiphase flowmeter? What methods of proving a multiphase meter could be used: - portable test separatorsystem; another multiphase meter; testing by difference using the production separatormeasurements or where the sum of individual well flows is balanced?

152.2 Framo Multiphase Metering System (JIP - BP a partner)

As for the BP multiphase meter, this system is based on a technique which should in principlebe insensitive to the non-reproducible nature of multiphase flow.

It comprises a novel mixer or flow conditioner which, unlike conventional mixers, can bedesigned to mix slugs and gas pockets and produce a steady exit stream from intermittentregimes (It, in effect, can mix phases segregated axially along the pipe). The flow conditionervessel should, in practice, be compact relative to a separator and its functioning is entirelypassive (no moving parts, valves or level controls).

The system is under development by Framo Engineering AS in Norway. It is at an earlier stagethan the BP meter and a prototype is only just approaching a first field trial. Sensors used in theconditioned stream immediately downstream of the mixer to measure the flow rates of oil, waterand gas are described below. Framo have conducted oil, water, gas flow loop tests at theirFusa Fjord dock-yard.


:I


Total flow

A venturi is being tested. Data to date indicate questionable accuracy. It is currently unclear towhat extent this is attributable to the challenge of measuring pressure differential accuratelyover a wide turn-down (the measurement data are particularly erroneous at high gas fraction). Itmay also be that, in spite of the mixing of the flow stream, localised phase slip effects stillpersist through the measurement section.

Phase fractions

Dual energy gamma ray attenuation measurements are used to determine the ratios of oil,water and gas. As for the total flow measurement, accuracy is currently subject to an unaccept-able degree of uncertainty. Again, the extent to which phase slip is responsible is unknown.Certainly, the current basic dual energy gamma ray detection system is subject to fundamentallimitations. However, new technology and steps within the JIP to improve the system could yetmove the system forward.

The most recent test results show relative measurement errors in phase flow rates and totalflow of 20% to 30%.

-

The mixing unit could, of course, ultimately be used with other measurement devices. Dielectriccomposition measurement techniques being developed elsewhere might do better than thedual energy gamma system. It is possible that the mixer could assist cross-correlation velocitymethods to achieve a degree of consistent and reproducible measurement.

The current JIP is developing this multiphase metering package as an offshore module. FramoEngineering’s expertise lies principally in this area. The cost to supply a Framo mixer basedsystem, either as an offshore package or in terms of the off-shelf elements of the multiphasemetering system, is unknown. The field prototype, designed for a nominal 3” or 4” line, is in abarrel approximately 4 m long by 0.5 m diameter. Its component parts, engineering and manu-facture resulted in a total cost of approximately $900,000. Total weight is 2.5 tonne. This unit isbuilt for top-sides use only, but some of the key sub-sea design of component parts (such assubsea housing of sensors and electronics processing and data transmission, barrier fluidsystem, electric coupler etc.) has been accommodated in the prototype. It is rated for class2500 pressure service.

The complete prototype system has exhibited pressure drops reaching 2.6 bar during the flowloop test programme. This was realised at flow rates comparable to the top of the range of theprototype BP multiphase flow meter.

Calibration of the Framo system would rest with the type of measurement used. For the venturimeter, it is primarily a matter of subjecting the pressure sensors to accurately known fluid pres-sure. This does assume that a universal or repeatable and predictable venturi discharge coeff-cient can be established for well mixed multiphase flow. Flow loop progress on this so far hasnot been good.

The dual energy gamma system requires calibration on fluids of known density and attenuationcoefficient. Once installed, some checking should be possible using a barrier fluid (pumped viaumbilical) system which Framo have incorporated into the test prototype design for variousfunctions such as sealing and pressure leg flushing. However, variations in phase properties,such as water salinity or attenuation coefficients of the phases with any change in phase



composition, would result in the need for some means of re-calibration in-situ. This could resultin a need to take fluid samples unless some means of separating and isolating individual phasesin the measurement section can be developed*.

* Such a technique has been demonstrated by the author for measuring individual phasedensity using the BP prototype multiphase flow meter at Wflch Farm and during the later trial atPrudhoe. Bay in 1992. It relies on the installation of the multiphase meter in a by-pass and blockvalve arrangement. The procedures would be part of the overall well testing procedure.

The question then remains as to how calibration will change with any service degradation of, forexample, the critical venturi throat and gamma beam path cross-stream dimensions. Here, asfor the BP meter, the desirability for regular re-calibration, possibly in-line using some means ofproving, must be addressed.

A further question arises as to whether the Framo mixer can be designed in any application toaccommodate the full range of flow regimes, flow rates and phase fractions expected over thelife of the well pad. (The same question applies to the venturi). Framo have developed a designbasis for the mixer. For example, it can be sized and dimensioned to suit a certain range of slugand bubble lengths. Framo claim, but have yet to prove, that it can be designed to cover the fullrange of regimes including annular. Whilst this may in principle be correct, it remains to beseen in practice whether universality can be achieved without excessive bulk and multistageconfiguration of the mixer.

15.2.3 The CMR (Christian Michelsen Research)/Fluenta Capacitance Cell(BP sponsorship 1985-1992)

This full-bore non-intrusive phase fraction sensor has been developed with significant assis-tance from BP alongside the BP multiphase flow meter. It measures in-situ phase fractionswithin a pipe cross-section where the phases are uniformly distributed. Absolute errors of +/-2% phase fraction have been demonstrated by the test programme for gas void fractions(GVFs) of up to 50%.

The existing cell will operate on oil-continuous liquid phase only. Water continuous liquid (or awater film around the sensor lining wall) results in a short circuit and meaningless reading. CMR(formerly CMI) are developing an additional sensor for the system which uses an inductanceprinciple to measure phase fractions with water-continuous emulsions. Both types of sensorrely on a simultaneous gamma ray attenuation measurement of cross-section mixture density inorder to resolve the three phase fractions.

The water cut may be derived from the phase fraction measurements. A class 1500 version ofthis device was tested downstream of the BP multiphase meter at Prudhoe Bay in 1992. Thesetests indicated that, by dynamically tracking the varying cross-sectional phase fractions in themultiphase stream, the cell could give accurate readings of water cut (within 1% absolute) forGVFs close to 90%. The limitation on achieving this in all cases was imposed by annular flowingwell streams and flows where slugs of significant duration and liquid hold-up (instantaneousGVF well below the stream overall average level) were not in evidence.

Calibration of the phase fraction cell for the field trial was completed at the CMR laboratory.Samples of dead crude and formation water were shipped from Prudhoe Bay to enable calibra-tion in a special purpose laboratory dielectric measurement cell. This is used to pressurise thesample oil and force into solution methane gas under varying pressures. As well as requiring as



input constants the dielectric properties of the individual production fluid phases, theCMR/Fluenta system also relies on knowledge of the densities of the phases as a function ofline pressure and temperature.

CMR and Fluenta have adapted the capacitance sensing section into a multi-electrode axialarray in order to yield velocity measurements by cross-correlation processing of the electrodesignals. The continuing project is now attempting the measurement of dual velocities by usingcombinations of different length electrodes in order to make some allowance for slip betweenthe gas and liquid phases. All the drawbacks highlighted in the earlier comments, concerningthe uncertainties associated with cross-correlation flow velocity measurement, apply. An accu-racy figure for phase flow rates measured using the combined phase fraction and cross-corre-lation measurements cannot, at present, be quoted with any meaning.

The 1992 test programme, including the BP field trial and quite separate test programmes bytwo other operators, exposed a hitherto not experienced fundamental flaw with the use of theceramic insulating liner material. (Development of a ceramic liner was pushed hard by CMRduring earlier development of the cell and was a significant part of it). It has been discoveredthat the ceramic has a surface affinity to water under certain flowing conditions which results ina completely erroneous sensor response signal.

Fluenta are investigating alternative liner materials. Further extended field testing will be requiredto establish the suitability of any new candidate insulating materials.

Typically, Fluenta have quoted on the order of $300,000 for one of their multiphase meteringsystems in the past, though recent remarks from CMR suggest that this quotation will changein view of the lower price being quoted for a competitor’s system (Kongsberg, see below). Thesensor is a compact spool, approximately 0.5 m in length and weighing one tonne in the caseof the cell tested at Prudhoe Bay. In view of the recent CMR remarks, $300,000 clearly doesnot relate to the basic manufacturing cost.

The CMR/Fluenta test and development programme continues in strength in spite of BP’s with-drawal from the JIP in 1993. It has support from several oil companies in three separate testprojects.

--15.2.4 The MFI (Multi Fluid Inc.) Microwave System (IIP - BP a partner)

A competitor of the CMWFluenta system, this microwave instrument also measures mixturedielectric constant and gamma ray attenuation (mixture density) to resolve the in-situ cross-sectional phase fractions of oil, water and gas.

In one configuration, the instrument’s microwave transmission and sensing section (whichmeasures the phase shift of the microwave energy caused by the multiphase mixture, related tothe phase proportions) functions in water continuous as well as oil continuous emulsions. Thisversion of the MFI instrument, known as the “full range meter” is still undergoing laboratory leveldevelopment of a field prototype. A first field trial is expected to take place during summer 1993at the Elf operated Pecorade site in France.

A lower price version (so called LP version, not to be confused with low pressure) has lent itselfto faster track development and has already been offshore on Statoil’s Gullfaks B platform inlate 1992. This configuration of the microwave sensing spool permits measurement with oilcontinuous emulsions only, but is more sensitive (and expected, therefore, to be more accurate)than the full range variant.



The ‘low price’ oil-continuous instrument is currently quoted at an estimated price of $200,000by MFI. This includes a twin sensor arrangement which permits a single velocity measurementby cross-correlation. A price for the more expensive full range device has not yet been quoted.The sensors are intrusive, using antennae probes which protrude into the flow. The LP sensorcomprises a resonant cavity, bounded by a grid of metal strips across the flow stream at eachend of the sensing spool section. The prototypes of this version have been constructed to class900 pressure rating. Further development is required in order to ruggedise the design of theprobes of the full range meter for multiphase flow stream service.

The sensor spools of the prototypes built to date are 3” bore. The two parts of the meter incor-porating cross-correlation measurement of velocity result in a measurement spool approxi-mately 1 m in length. One flow composition measurement spool accounts for half this length. Afurther spool housing pressure and temperature transmitters is necessary (this also applies tothe FluentaKMR system).

The Gullfaks B trials of the LP meter have resulted in only a few tens of hours exposure toproduction fluids so far. The phase fraction data agreed well with test separator measurementsfor the limited number of test cases. The GVF at the Gullfaks B test site is below 40% and theevidence suggests that the multiphase stream is well mixed and steady in nature.

Thus, these tests represent the easy end of the multiphase spectrum. Phase fraction resultswere within the +/- 2% absolute errors seen in laboratory flow loop tests. The few cross-corre-lation measurement data, in terms of both mixture velocity and phase flow rates, were moreerratic. Relative errors of +/- 10% to 20% were reported.

The steady and relatively uniform nature of the low GVF multiphase stream of these field trialsshould have provided a flow structure and velocity profile of minimal complexity and leastremoved from that of the flow loop testing by MFI. This should have favoured the cross-correla-tion technique. However, such a flow stream also exhibits relatively weak perturbations inmixture dielectric properties on which to cross-correlate. This would tend to give a broad andrelatively ill-defined correlation peak from which to deduce time of flight (to deduce velocity).

As for the CMWFluenta cell, calibration of the MFI phase fraction system, in principle, shouldrely primarily on factory or laboratory set-up prior to installation. Both systems use commerciallyavailable Cs137 gamma ray densitometers, as used for the BP multiphase meter, which can becalibrated on fluids such as air and tap water. The microwave unit requires data includingdensity of the dead crude, produced water density and conductivity. The MFI system then relieson application specific prediction equations of the variation of these quantities and gas densityas a function of line pressure and temperature.MFI can configure their flow composition software to derive the flow stream as a split of totalmass proportion of hydrocarbons and water, as opposed to volume fractions of oil, water andgas. These output readings are far less sensitive to uncertainties in the individual phase densi-ties than the outputs of the phase fractions.

This split of output data would, of course, need to be related to total flow throughput to derivethe total hydrocarbon and water mass flow rates. This could only be achieved if it could alwaysbe ensured that the phases are well mixed and flowing all at one velocity through the measure-ment spool (zero slip).

The question needs to be answered as to whether output of total hydrocarbon mass and waterthroughput is a more useful operating measure than in-situ volume flow rates of oil, water andgas. The latter usually require the application of correlations to account for shrinkage and solu-


Section 15. Meterha BP Multiphase Design Manual

tion gas factors in order to convert to flow rate units expressed at standard pressure andtemperature.

Like the phase fraction cell developed by CMR, the fraction part of the MFI microwave tech-nique could be used to supply a water cut reading to the BP multiphase meter system. It wouldbe used to dynamically track the phase fractions and should yield water cut in high GVF flowsas well as covering the lower part of the GVF range. Such an off-shelf option should presum-ably cost less than the estimate given above for the MFI LP flow meter, which includes an extrasensor and software to allow cross-correlation measurement of velocity.

It should be noted that MFI have already successfully commercialised a water cut sensor usingthe same microwave hardware configuration as the LP meter. This instrument is for gas-freeemulsions and should not be confused with the multiphase flow meter technology describedabove, which incorporates a gamma densitometer and additional sophistication in order tomeasure oil-water-gas flows.

15.2.5 Neural Networks by EDS-Scicon and CALtec Ltd (JIP - BP a partner)

The current JIP started in 1992. The aim of the project is to develop a multiphase meteringtechnique which uses advanced parallel processing - neural networks - to apply pattern recog-nition to the dynamic signals from relatively simple, low cost instrumentation such as gammaray densitometers and pressure transmitters.

Neural networks have the capability to “train” or “teach” themselves. to recognise “features”from complex phenomena. The idea is to “train” the neural network to relate “recognisedfeatures” to the flow rates of oil, water and gas.

Activity, so far, has involved “training” different structures of neural network on flow loop multi-phase flow data and some gamma ray densitometer field data (Forties Echo-Alpha pipelines)supplied by the multiphase flow group within BP. The object was to give EDS-Scicon theopportunity to work the networks on multiphase flow and demonstrate the technique. Typically,this has involved using a data set of 40 flow rates and predicting each data point (oil, water andgas flow rates) by training on the other 39.

Flow rate relative error data produced by this exercise exhibit uncertainties broadly in the rangeof 10% to 30%. However, the work is at an early stage. One of the difficulties facing theapproach is the general lack of abundant, reliable and accurate field data containing oil, waterand gas flow rates and corresponding signals from sensors such as pressure transmitters.

The programme will need to address which types of sensor are most suitable in detecting char-acteristic features, and where on the pipeline they should be installed to pick these out.

The method requires specialist expertise to adapt and set-up the neural network software tothe application. The cost of this and of the computing capability has not yet determined.

It works best where there is an abundance of data on which to train the neural network andwhen interpolating between these data but not extrapolating. Given the general non-repro-ducibility of multiphase flow this may make calibration of the system as a universal or stand-alone metering solution impossible in practice.


--


If the neural network approach can be provided at low cost, it could lend itself to applicationswhere the system would be trained against other expensive equipment (other multiphasemeters or test separator systems) where this has limited availability and therefore impedes thewell testing and reservoir management capability.

15.2.6 Texaco Starcut Meter

This water cut monitor has been developed by Texaco EPTD. Jiskoot Autocontrol Ltd. nowhave the exclusive world-wide licence to manufacture and market the product. It usesmeasurements of the phase shift and attenuation of 10 GHz microwaves caused by the flowstream to determine water cut.

The instrument will measure water cut of emulsions containing little or no free gas. It works withoil-continuous and water-continuous emulsions, i.e. water cut from 0% to 100%. JiskootAutocontrol can quote a number of oil industry site applications and state that gas volume frac-tions of up to 25% can be handled. The instrument is already developing a field track recordahead of the other phase fraction/water cut devices described above, although these instru-ments have, of course, already demonstrated a measurement capability on gas fractions signifi-cantly higher than 25%.

Texaco’s development effort has concentrated on accumulating a vast data base of “mixingcurves” for a wide range of oils and water chemistries. This is stored in the system’s micro-processor, packaged with the sensor in a compact unit. The unit can continuously perform anauto-calibration in field service to ensure the correct mixing curve (the curve which tracesoil/water ratio from the microwave phase shift and attenuation measurements).

The detection speed of the device is quoted as >300 samples/second It is claimed that it cancope with variations in fluid properties such as salinity. This could give it a significant advantageover the CMWFluenta and MFI systems which rely on operator input of salinity, which thereforemust be fairly stable, and equations to allow for its variation with temperature.

The Starcut meter determines phase properties such as crude oil SG or density and watersalinity and provides output signals which track these properties. The ability to track salinity isan attractive option for reservoir management. It could provide an indication of water floodprogress or potential problems with premature breakthrough of injection fluids, for example.

Texaco claim the Starcut device can tolerate up to 25% GVF in the flow stream at present,though no data have been published to substantiate this. They state that development workcontinues in order to increase the free gas content acceptable and the capability to measurethe gas fraction.

The sensing flow path is of rectangular cross-section of approximately 10 mm by 5 mm normalto the flow path. This limits the device to rely on a slip stream sample arrangement in themajority of applications since it clearly imposes a throughput limitation (quoted at 200 bpdliquid).

Should the capability to measure water cut with free gas presence (i.e. in multiphase flow)improve to a sufficiently high GVF, the Starcut cell is an attractive option for installation as a slipstream unit in combination with the BP multiphase flow meter. It would be used to provide therequired water cut reading. It would readily fit via small bore piping (perhaps 20 mm to 25 mm



diameter) as a compact side unit. Furthermore, the unit is available to the highest pressureratings.

The present Jiskoot pricing for the off-shelf unit, of approximately $70,000, would also makethe Starcut unit a preferable option for the BP meter relative to the MFI and Fluenta measure-ment systems.

The throughput limitation of the Starcut monitor and its consequent limitation as a slip streamdevice could be viewed as disadvantageous relative to the Fluenta and MFI full bore measure-ment cells. The test work with the BP screw meter has indicated that it enhances mixing of thewater and oil (this has been demonstrated by taking side stream samples from the meter manu-ally during flow loop tests). The combination of the screw meter with the Starcut monitor wouldbe expected to provide the latter with a well mixed and representative sample from which toderive water cut.

Calibration of the Starcut monitor is essentially a factory/laboratory exercise. It subsequentlyperforms its own auto-calibration once installed and in service.

15.2.7 Kong&erg Offshore AS- Capacitor Cross-Correlation Multiphase Flow Meter

In 1992, Kongsberg Offshore commercialised a multi-capacitor multiphase meter developed bythe Shell Exploration and Production laboratory (KSEPL) in the Netherlands.

The measurement spool comprises two closely spaced (10 mm apart) parallel plates mountedacross the centre of the pipe. The spool is mounted in horizontal pipeline with the plates verticaland in the plane of minimum resistance to the flow stream (i.e. aligned in the direction of flow).The plates support a column of small rectangular electrode pairs which spans the diameter ofthe pipe. Two more electrodes are positioned further downstream on the plate, one near thetop and one close to the bottom of the pipe.

The meter has been designed to measure flow rate of oil, water and gas in slug flow. The gas-liquid interface can be sensed by the electrodes. The flow perturbations sensed at the upperand lower downstream electrodes are cross-correlated with the corresponding signals of theupstream electrodes.

Simultaneously, the capacitor outputs are used to determine the cross-sectional area occupiedby gas and liquid in the pipe by measuring the liquid hold-up. This information is combined withthe cross-correlation velocity measurements to determine gas and liquid flow rate. The tech-nique processes the electrode signals in such a way as to make some allowance for theproportion of gas entrained in the main body of the liquid.

The modelling assumes a slug velocity equal to the gas pocket bulk velocity. Water cut is deter-mined from capacitance sensed by the lower electrodes immersed in the liquid film of the slugflow. Currently, the unit only works for oil-continuous emulsions.

Shell state that units have been undergoing field trials in Oman and Gabon with “consistent”results compared to test separator equipment although exhibiting some bias error. They intendto continue funding development to extend the capability of the system to operate in non-inter-mittent flows and water-continuous emulsions (though they admit the latter could proveunfruitful). Low pressure flow loop results published show gross errors in flow rate measure-ment outside of a cluster of points lying within the intermittent region of the flow regime map.

-



Shell’s intention has been to target this type of meter at remote land-based sites with significantpipeline between wellhead and manifold where well defined slug flow occurs. They claim tohave a number of such applications and state that, in these cases, the level of accuracy achiev-able with the unit (quoted at between 10% to 20% of flow rate) is acceptable.

In sacrificing accuracy by the nature of the technique, Shell have attempted to evolve a rela-tively low cost unit, which could allow one meter per well. (Less than $100,000 unit price,including industrial PC, has been quoted with a target for Kongsberg to reduce this to $50,000during the course of the collaboration between the two companies). Also, Shell wished to avoiddeployment of nucleonic sources (common to most of the other techniques) at remote locationson land.

Currently, the units are available in 3” and 4” line sizes to ANSI class 600 rating. The 4” sensorspool weighs 90 kg and is 556 mm in length. The Kongsberg data sheets do not include a sourservice specification at present.

In principle, the cross-correlation measurements to measure velocity should not rely on flowcalibration. However, as noted above, Shell have reported some systematic errors during fieldtesting. The compositional calibration of the capacitors relies on readings with the separatephases present. In this respect, the method has similar set-up requirements to theFluentaICMR system which also, in effect, relies on knowledge of the capacitance reading onsingle phase fluid.

152.8 Further Comments

The above review of multiphase metering techniques covers the most advanced equipment ofwhich the author has technical knowledge. There are other developments of which the multi-phase flow group is aware.

Schlumberger (Cambridge Research) are believed to be working with a multi-electrode capac-itor technique, where rotating electrical fields are applied to multiple electrodes positionedcircumferentially around the pipe. This type of system attempts to work with non-uniform spatialdistribution of the phases within the pipe cross-section. It has been investigated elsewhere also.

.Arco are known to have tested a system relying on separation of the gas and liquid phases. Co-operation between the multiphase flow group and Arco in conducting the Prudhoe Bay fieldtrials of the BP multiphase metering system may reveal more details of this.

Texaco also have worked on a much publicised separator system. This, so called, “multiphasemeter” comprises a 10m high structure weighing many tonnes, which houses an inclined sepa-rator system and associated pipework. Vortex shedding and variable area single phasemetering systems are used in the gas and liquid legs respectively off the separator. Texacomicrowave technology is used in a liquid leg to measure water cut. The projected cost of theunit when publicised in 1990 was up to $3 million. Field trials on the Tartan-Highlander subseatie-back system in the North Sea met with difficulties such as associated with foaming of themultiphase stream. The development work by Texaco and Jiskoot Autocontrol was continuingand the package was scheduled for installation on the subsea Highlander template in 1992.The author is not aware of any further publicity since 1991.

A downhole venturi meter has been developed jointly by Aberdeen-based company Exal andBP. This device cannot tolerate the presence of any gas and presumably relies on a uniform



and well mixed emulsion. A prototype was commissioned in an ESP well at Wytch Farm inJanuary 1993. The device measures total liquid flow rate and future installations are expectedalso to yield a water cut reading by head measurement of the liquid column. At the time ofwriting the author has still to establish details of the Exal development programme and funda-mental basis (for example, oil-water flow regime behaviour downhole). Such equipment couldprovide the most suitable means of measuring well fluid rates in certain applications. It mightalso be used as part of overall multiphase metering schemes in combination with other hard-ware under development.

Several other multiphase meter development programmes are or have been in existence. Forexample, the Euromatic combined mixing/side-stream separator system reached and under-went a field trial stage but seems to have faltered owing to its significant lack of sound funda-mental basis and understanding of multiphase flow behaviour. The Mixmeter JIP programmelaunched by Imperial College, London, in 1992 and combining some elements of the basictechnology already covered in the JlPs listed above, is an attempt which might bring newoptions to the fore in the future. The purpose of this review is to describe in some detail what isconsidered to be the most soundly based and advanced technology to date. Not all of theknown approaches have been listed.

Most, if not all, of the techniques covered in the above review currently have a high degree ofmeasurement uncertainty at extreme GVF (95% and above). The total volumetric measurementof the BP screw meter has been shown to be the exception to this. Annular liquid flow patternsin the rotors are believed to assist in sealing the clearances. (Also, the venturi meter is known tobe used by Shell for measuring total volume flow rate in gas condensate well flows and ultra-sonic wet gas meters have been developed. However, for these two cases, gas flow rate butnot liquid rates in gas condensate streams are measured).

Flow patterns which are predominantly annular in nature, which can persist at lower than 90%GVF as line pressure increases, currently cannot be handled by the Fluenta/CMR and MFIphase fraction measurement systems (or, indeed, the Shell/Kongsberg meter). For these instru-ments, some form of flow conditioning to perturb the annular flow pattern would be neededwhere such conditions are expected to exist during the service life of the flow meter.


BP Mu&phase Design Manual Section 15. Metering

15.3 summaryThe main techniques for multiphase metering which have emerged can be broadly classifiedinto two categories. One type of approach employs flow control or conditioning in order toestablish reproducible conditions at the measurement section. Other techniques have adoptedthe principle of cross-correlation velocity measurement in combination with phase fractionmeasurements to establish the phase flow rates. This latter approach is highly susceptible tothe complex and generally non-reproducible nature of multiphase flows.

The general recommendation of the philosophy is that those multiphase metering techniquesemploying flow pattern-independent principles, that is, the former of the above two approach-es, will most readily facilitate a calibration methodology and, therefore, be of use. For certaintypes of these technologies, such as the BP screw meter based system, test experience isdemonstrating a way forward using relatively straightforward and low cost factory calibrationmethods utilising single phase fluids. All measurement techniques are likely to require wide testexperience to evaluate and evolve standards. The level of accuracy and its qualification requiredwithin the oil industry will dictate the extent of the testing effort. Please see section 16 forfurther discussion of the calibration of multiphase metering systems.


Section 15. Metertng BP Multiphase Design Manual


LEFT BLANK



(1)

(2)

References

‘Study to Install a Multiphase Meter Subsea, BP Engineering’,BP Engineering Technical Note, ESR.93.ER.036.

‘Proposal to Install a Multiphase Meter Subsea on Cyrus’,BP Engineering Technical Note, ESR.93.ER.035.

-



-


LEFT BLANK


BP Multiphase Design Manual Section 16 Meter Calibration

Section 16. Meter Calibration

Calibration of Multiphase Meter Systems

16.1 Introduction

16.2 Aspects of Flow Calibration

16.3 Multiphase Flow and Calibration

16.4 Multiphase Measurement Standards in Practice

16.5 Field Calibration or Proving


Section 16 Meter Calibration BP Multiphase Design Manual

--


LEFT BLANK

-


BP Multiphase Design Manual Section 16. Meter Calibration

16.1 Introduction

In this section, the basic aspects associated with calibration of multiphase flow meters arelisted. The special challenges posed by multiphase flow - in particular, the difficulty of ensuringreproducibility, without which a calibration cannot exist - are then highlighted. In the face ofthese challenges, some theories, based on the BP test experience of metering multiphase flowsto date, are presented as to how multiphase flow meter calibration philosophy might develop inpractice.




LEFT BLANK


BP Multiphase Design Manual Section 16 Meter Calibration

16.2

16.2.1

Aspects of Flow Calibration

Reasons for measurement

The multiphase metering systems currently under development have been targeted at well flowmeasurement for reservoir management and allocation. The significantly more stringent accu-racy usually associated with custody transfer or “fiscal” flow metering has not been addressed.

Clearly, any means of calibration of a multiphase flow meter should be appropriate to the levelof accuracy required. As a general rule, the reference system should be prone to significantlysmaller uncertainty levels then the meter under calibration. A factor of 10 superiority of uncer-tainty levels of the reference meter might typically be desirable in some conventional singlephase calibration scenarios.

16.2.2 Traceability

Strictly, the measurement uncertainties of the reference metering should be qualified by primarycalibration against a recognised standard. This will take place at regular intervals dependent onthe circumstances.

16.2.3 Proving

The specific case of a reference calibration or master meter checking system installed in bypasspipework at the metering station which can periodically be switched to measure flow and checkor calibrate the flow metering system in-situ. The proving system can be a fixed installation orbe a portable servicing system.

16.2.4 Repeatability

Repeatability is a pre-requisite to flow meter accuracy and is determined by means of a speci-tied number of repeat comparisons against a reference measurement. This procedure relies onrepeatability of flow conditions (and, of course, of the reference measurement).

A calibration based on a flow loop test, such as by the manufacturer of the flow meter, relies onthe concept of a repeatable or predictable and reproducible flow condition.

For example, the requirement for specified pipe straight lengths and flow straighteners inturbine meter installations ensures known calibration flow conditions are re-produced in singlephase flows. Indeed, single phase pipe flow is well enough understood and its behaviour suff-ciently predictable that an orifice plate can be manufactured to controlled tolerances such thatflow calibration is not needed. The accuracy of the orifice plate can then be traceable bymetrology


ISection I6 Meter Calibration BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 16 Meter Calibratton

16.3 Multiphase Flow and Calibration

In measurement, the concepts of repeatability and traceability are the foundations of calibration.

Multiphase flows are more complex and less well understood than single phase flows. They aregenerally not predictable in the same way as single phase pipe flows. For example, multiphaseflow behaviour cannot be guaranteed in terms of a steady and defined velocity profile even ifcertain rules regarding pipe length are obeyed. In general, the multiphase flow behaviourdepends on the entire history of flow path from source and upon influences from a significantdistance downstream. Experience has shown that the nature of the multiphase flow patterns isinstallation dependent and therefore not necessarily repeatable or reproducible between oneinstallation and another. A manufacturer’s multiphase flow loop calibration would not, ingeneral, necessarily represent the field installation.

Even for the case of a proving system, the change in flow path in bringing on line the provercould alter the multiphase flow behaviour at the flow meter and, therefore, its response charac-teristic.

Clearly, the foundation of repeatability or reproducibility is much more difficult to realise in multi-phase flow.

In addition to the installation dependency of the multiphase flow behaviour, the nature of flowpattern can drastically change through the service life of the flowline. For example, a multiphaseflow meter might be subjected to a stream of mainly oil flow containing dispersed gas bubblesin the initial stages of service. As gas content increases with the depletion of the reservoir, themeter could be subjected to intermittent flow behaviour, such as plug or slug flow. Finally, in theextreme case, the meter could have to cope with annular flows of high gas content and signifi-cantly higher total fluid volume throughput than in earlier service life. Additionally, the latterstages of reservoir depletion and any associated secondary and tertiary recovery schemescould result in an increasing portion of water in the flow stream.

Thus, in terms of calibration (quite apart from meter performance) multiphase flow metering canpose significant new demands of flow pattern behavioural change and flow rate range (turn-down).

The above discusses calibration for multiphase flow meters by relating to the case of singlephase flow. It has highlighted the difficulty in realising reproducibility in multiphase flows todetermine measurement response. But, of course, measurement response in multiphase flowrelates to more than one response, or signal, which must be related to more than one (three -for oil, water and gas) flow rate. The calibration process in multiphase flow, then, can be evenmore involved than for existing single phase methods from this consideration also.

The multiphase meter measurement can involve several simultaneous sensor responses. Thecombined influence of the errors in all these signals will determine the overall accuracy of themultiphase meter in measuring the phase flow rates and this must be determined by calibrationprocedure.

Further, most, if not all, of the multiphase metering techniques under development rely onknowledge of the individual phase properties, such as oil, water and gas densities, in order toderive the phase flow rates from the sensor inputs. The calibration and overall specification ofaccuracy and repeatability for any multiphase metering installation must include an account ofthe uncertainties associated with the measurement or prediction of these properties.


Section I6 Meter Calibration BP Multiphase Design Manual

The greatest cost benefits attributable to multiphase metering are expected to be realisedwhere the new technology is installed subsea at well pads remote from the processing facility.The requirement for remoteness subsea poses another significant challenge in terms of calibra-tion.

How will it be ensured that a meter remains on calibration specification?

This is important not just in terms of deciding upon the level of calibration as necessary routine.The possibility for major changes in the nature and throughput of the flow stream, as discussedabove, could conceivably dictate a requirement for calibration checking or proving on-line. Isproving of a remote meter in a multiphase pipeline practical? Would such a requirement gosome way towards negating the advantages of a multiphase metering scheme?

-


BP Multiphase Design Manual Section It5 Meter Calibration

16.4 Multiphase Measurement Standards in Practice

From the above, it can be concluded that, just as conventional single phase meters cannot beapplied to multiphase flow measurement, calibration methods based on single phase flowunderstanding cannot be relied upon.

New or modified approaches of calibration need to be adopted for multiphase metering.

It could be argued that, given the general non-reproducibility of flow behaviour in multiphaseinstallations, measurement techniques exhibiting minimal sensitivity to multiphase flow patternbehaviour will not only lend themselves to calibration, but may also provide the means todefining calibration technique and reference standards.

This would be achieved on the basis of proof by experience and physical argument. The degreeof rigor involved in this, and hence the level of development and time to evolve a calibrationmethodology, will depend on the flow measurement uncertainty levels acceptable and theextent of qualification required by the oil companies and government agencies. These partieswill need to take an involved and informed view of what the technology developers and manu-facturers are able to provide.

Some multiphase metering systems under development such as the BP patented screw meterand Framo static mixer based system (see Section 15 - Multiphase Meter Systems and TheirApplication) involve techniques which attempt to minimise flow pattern dependency ofmeasurement by controlling the flow behaviour. Only experience of tests in several installationsof widely varying flow conditions, involving comparison against accurate single phase refer-ences downstream of separators, will develop and confirm the uncertainty levels achievable andthe universality of the calibration of these approaches. The alternative to protracted testing is forall concerned (oil producers, government agencies, manufacturers and technologists) to acceptargument based on sound theoretical principles with only limited flow loop and field testevidence of proof of theory. Given its physical complexity and comparative youth as a branch offluid dynamics tackled by engineering science, multiphase flow will not readily lend itself to thelatter approach

Other multiphase metering approaches have addressed measurement using sensors whichprimarily respond to the composition of the fluid stream. These development programmes haveconcentrated on ways of measuring the phase fractions or ratios but latterly some haveadapted the sensing elements to yield additional cross-correlation measurements of fluidvelocity.

The calibration of such sensors to measure mixture composition is a relatively straightforwardproposition where it can be ensured that the phases will be well mixed or uniformly distributedwithin the sensing volume in service. Typically, the techniques for measuring phase fractions,described in Section 15, apply mixing models relating sensor response to mixture composition.The calibration curves by these models are fixed by measuring the response signal on thesingle phase fluids.

The tendency of multiphase flow to form segregated patterns of the phases in the pipe caninvalidate calibrations by this approach. Where the sensors have a fast response, it is possiblefor these devices to track fluctuating flow stream composition provided the phases are distrib-uted uniformly in the cross-sectional plane. This can be achieved by mounting the compositionsensor in a vertical spool although some form of flow conditioner may still be required to breakup annular flow (or intermittent annular flow pattern) when this establishes upstream of the



sensor.Some highly sophisticated types of multi-sensor compositional devices are being investigated.These systems are being developed to have a rapid response which accounts for phase distrib-ution within the sensing zone. The complexity of these approaches may make them difficult toprove in practice.

If composition can be accurately dynamically tracked, it must be combined with simultaneoustotal flow rate or velocity measurement to provide phase flow rates. Even in dispersed oruniformly distributed flow patterns, slip (the difference in velocity between the phases) can besignificant. Thus, either some form of flow conditioning to eliminate (or minimise) slip is requiredor else the individual in-situ phase velocities must be measured.

Cross-correlation measurement of velocity has the potential in principle to be “absolute”. Thatis, flow calibration is not necessary. Accurate measurement of sensor spacing and “time offlight” will give accurate measurement of velocity.

However, the accuracy of the “time of flight” measurement as derived from a cross-correlationfunction, or more specifically, what it actually represents, is greatly influenced by flow velocityprofile, turbulence and signal noise. In multiphase flow, as discussed already, velocity profile isgenerally highly complex and non-repeatable in nature and slip generally occurs between thephases. Accurate cross-correlation measurement would rely on a flat cross-sectional velocityprofile or multi-velocity measurement of sufficient detail to allow for profile shape.

What is more, mixture property cross-correlation devices tend to sense particular features ofthe fluctuating flow pattern. Some techniques now being developed are sensitive to the inter-face between liquid continuous bulk and large gas pockets. Others use sensing fields which aremost sensitive to certain regions of the pipe cross-section such as the near-wall. It does not,therefore, follow that these cross-correlation techniques measure bulk (average) flow velocitywhich can be multiplied by cross-sectional area (adding further uncertainty) to give flow rate.

At least two projects are under way to measure more than one cross-correlation feature inorder to derive multiple velocity information which relates to the different phases. This is at anearly stage of development but could prove impractical to qualify as universally field applicable.

The BP screw meter and Framo mixer approaches of flow pattern independence by flowcontrol, combine a total flow measurement method with a mixture phase ratio measurement.This can allow total flow throughput measurement and phase fraction measurements to becombined to yield phase flow rates. The objective, as already stated, is to overcome the multi-phase flow characteristics of generally non-reproducible flows, complex velocity pattern andphase slip. Within the BP in-house development programme, this has been achieved bycombining the BP meter total flow measurements with a water cut reading derived from in-situphase fraction measurement.

If such flow pattern-independent techniques can be developed to an acceptable degree (i.e.acceptably repeatable signal responses for given flow rates in all applications) then flow calibra-tion on a manufacturer’s or standards institution multiphase flow loop (oil, water, gas) will bemeaningful.

Indeed, if such metering techniques are sufficiently successful then certain types of total flowrate meter could be calibrated on single phase fluids, such as water, since flow pattern inde-pendent response in multiphase flow is, by its nature, an insensitivity or predictable response tothe density and viscosity of the fluids being measured.


BP Multiphase Design Manual Section I6 Meter Calibration

Given the general necessity of only needing to calibrate mixture phase fraction instruments onthe basis of single phase fluid properties, the combined metering system could then be cali-brated at the factory or laboratory site, using relatively simple and established procedures andsingle phase fluids.

This type of approach will result in lower cost multiphase metering solutions than calibrationrequiring the use of relatively complex and expensive oil, water, gas flow loops. (For example,one particular multiphase flow meter test loop built by another oil company in Europe cost onthe order of $1.6 million to build in the mid 1980s. It is a low pressure, 4” nominal bore loopusing gas-oil, tap water and air which are metered to traceable standards. This requires regularprimary calibration of the multi-spool reference metering and six monthly checking of theproving system by the national standards authority. Additionally, full-time maintenance back-upis employed. This degree of rigor is necessary in order to qualify multiphase metering to +/-5%relative errors; - a target usually quoted for multiphase flow meters).

Of course, after initial set-up, calibration and installation, there remains the question of subse-quent calibration or proving in service. Further thought on this is prompted in the review ofmultiphase metering techniques in Section 1.5. In that section, questions relating to the practi-cality of certain field proving approaches are raised. The benefit of ROC experience and inputwill be necessary in order to tackle the proving issue in particular. Ultimately, the provingrequirement of the flow pattern-independent techniques could evolve to a basic requirementwhere simple static, dimensional or self diagnostic checks will ensure measurement to withinspecified uncertainty bands.

In summary of the above, two key approaches to multiphase metering have emerged. Oneapproach involves conditioning or controlling the flow stream to ensure reproducible flow condi-tions at the measurement section. The second approach uses the cross-correlation techniquein combination with flow composition sensors to derive the phase flow rates but is highlysusceptible to the complex and generally non-reproducible nature of multiphase flows. It couldbe argued that multiphase metering systems based on flow pattern-independent principles (thefirst of the above approaches) will most lend themselves to calibration and, therefore, useful-ness. Indeed, such techniques may well permit simple factory calibration and set-up proce-dures. Ultimately, these techniques could rely on only simple methods of field calibrationchecking which do not necessitate comparisons against flow measurements by another“master” meter or proving system. All measurement techniques are likely to require wide testexperience to evaluate and evolve standards. The level of accuracy and its qualification requiredwithin the oil industry will dictate the extent of the testing effort.


Section I6. Meter Calibration BP Multiphase Design Manual

-


LEFT BLANK


BP Multiphase Design Manual Section I6 Meter Calibration

16.5 Field Calibration or Proving

Certain questions, concerning the proving or in-service calibration of multiphase meters inparticular, have been raised in the preceding paragraphs. All of the emerging multiphasemetering techniques are likely to need some level of in-service checking, even if this onlyinvolves an initial, and subsequently infrequent, comparison against a reference measurement.

In relation to these questions, it is recommended that consideration be given to the implicationsof the following approaches to the problem of field calibration for practicality and accuracy.

16.5.1 Production Train Reference

Testing using the production train single phase metering system as a reference. This couldrequire that operations accommodate well shut-in periods, possibly flowing only one well at atime. Alternatively, it could involve testing by difference where one (or more) well is shut in at atime.

A more acceptable variation on this theme might be possible, with some types of multiphaseflow meter at least. Experience with the BP multiphase flow meter at Prudhoe Bay, for example,demonstrated repeatability for all the wells tested at the well pad. These all produced within atight band of gas volume fraction (although water cut and flow rate and behaviour variedmarkedly between wells). In such circumstances, the multiphase meter measurement wouldsimply be checked by comparing the sum of the individual well measurements with the totalproduction of the well site.

In the case of the BP multiphase meter, it can be assessed in terms of the total mass flow ratemeasurement of the stream flow rates or indeed to make an allowance for mass transferbetween the phases and changes in phase volumes from multiphase to reference measure-ment station.

Such a method, based on summing the production of the wells, would have to account for anydifference in the flow rate from the wells when switched between the production line and themultiphase meter by-pass, in cases where multiphase metering is not available on every indi-vidual well. Some multiphase metering hardware introduces a pressure drop (see Section 15.2covering the BP and Framo multiphase metering systems in the review of techniques).

A further alternative, which could be considered in certain circumstances, would be to flowmore than one well through the multiphase meter and compare its measurement directly withthe production separator metering system receiving the combined flow from the same wells.Clearly, this approach could be constrained by the maximum flow rate or turn-down of themultiphase meter, which would dictate that some wells be shut in during the proving exercise.

16.52 Portable Test Separator

Although involving additional high cost, particularly off-shore, another approach to consider isthe use of portable test separation equipment. This might be an option for multiphase meteringequipment requiring infrequent checking in certain circumstances. In addition to reservationsregarding the cost penalties of such an operation, this type of reference will be subject touncertainty levels of an order associated with the multiphase meter itself. It would serve thepurpose of a check rather than as a means to calibration.


Section I6 Meter Calibration BP Multiphase Design Manual

16.5.3 Multiphase Proving

This might consider adapting existing multiphase (or single phase - such as compact sweptpiston devices) metering technology for portable and non-continuous use. Certain types ofmultiphase metering technology, designed for prolonged service in extreme conditions of flow,might be made more cheaply for this purpose. It could also be manufactured and operated formore accurate performance. This would be possible with a switch of design emphasis fromrobustness in favour of accuracy. It would be realisable with short service exposure intervals,coupled with regular inspection and maintenance and operational precautions not possible inthe running of the installed multiphase metering system.

Using the BP screw meter again as an example, the clearances between the rotors and stator(approximately 1 mm) have been made sufficiently large to ensure a high degree of reliability inthe presence of particulates. It would be possible to improve the current accuracy of thissystem by reducing the clearances. Furthermore, limited production fluid exposure, avoidanceof proving coincident with certain well and reservoir workovers or operations, and the use of aserviceable fine gauze upstream of a meter in a proving skid would permit a design typical ofhigh accuracy fiscal applications of total volume flow rate measurement in single phase liquidflows.

-

16.5.4 Concluding Comment

It will be recognised that the acceptability of certain approaches could differ between land orsurface and subsea installations. Any further suggestions on field calibration are invited.


BP Multiphase Design Manual Section 17. Pressure Boosting

Section 17 Pressure BoostingMultiphase Pump Systems and their Applications

17.1 Introduction

17.2 Pressure Boosting- Methodology of Selection ‘BOAST’

17.3 Outline Review of Multiphase Pumping Systems

17.4 ‘Known’ Performance Track Record

17.5 Outline Design


Section 17, Pressure Boosting BP Multiphase Design Manual


LEFT BLANK

.-

-


-


17.1 Introduction

If the pressure within the reservoir is inadequate to transport the hydrocarbon fluids at therequired rate to a suitable place for processing them, then some form of pressure boosting willbe required. This may be by enhancing the pressure of the reservoir itself, eg. by pressurisedaquifer support and /or gas re-injection, pressurising the reservior from above. An alternative isthe use of down-hole single phase pumping, which does not affect the reservoir itself, but rais-es the well head flowing pressure. Lastly boosting may be applied downstream of the wellhead where the flow is normally multiphase. This Section considers these options and in partic-ular the use of multiphase pumping option.

This pressure enhancement may be required early in the field life or may only be requiredtowards the end of field life when the reservoir pressure has fallen too low for the standardmethods of transport to be adequate; never-the-less consideration of later means of pressureboosting should be considered in the ‘whole-life-field’ concept design. There are several formsin which this pressure boosting can be applied. Section 17.2 discusses the selection of themethod, including the gas-lifting of wells (which is really a means of pressure drop reductionrather than pressure boosting). Further consideration is given to one of these means of pres-sure enhancement - the application of multiphase pumping systems, see Section 17.3.

Multiphase pumps generally have higher power requirements than those of single phasemachines (pumps or compressors) performing similar duties. The theoretical power require-ment depends both upon the duty conditions and the theory used in calculation. As a resultattempts to compare performance in terms of efficiency tend to be misleading and are bestavoided.

Although they are described as ‘pumps’ it would be more accurate to describe these multi-phase pressure boosters as wet gas compressors, in that volumetrically gas is almost alwaysthe dominant phase. The volumetric throughput, that is the volume of the feed fluids at thefeed conditions, is the most important factor in pump design; in addition the gas volume frac-tion (GVF), the feed pressure and the pressure ratio are required.

Table 1 outlines the capabilities of the pump types which presently seem to have some com-mercial potential.


Multiphase Pump Types (Only Systems with established commercial potential are included)

Pump type Status Subsea ?Power source

Electrical HydraulicSand

ToleranceGeneral performance characteristics

Flow Delta P. GVF

Rotodynamic ~~ ~ ~~~~ ~~Multistage Helico-axial

Multistage CentrifugalContra-rotating axial (CRA)

P@tive DigplacementTwin ScrewReciprocating Piston (MEPS)Hydrobooster

OthersJet pump

‘With Separation’VASPSKBS

Key:Status:

Notes:

f Yes YesD Yes Yes

P/L Yes YesFr ? Yes

f No Yes NoFr Yes Yes ?P/L Yes Yes No

WI yes No Yes

P/LP/L

YesYes

YesYes

f = CommercialD = Commercial designs availableFT = Field testedP/L = Prototype and/or Lab. Tested

Yes?

Yes?

YesYesYesNo

MHLM

LimitedNo?

MM?

No L

NoNo

{4} As WELLCOM system(5) No Delta P for Gas

YesLimited

MI31H

Flow/DP/GVF HFlow {2)

m3/h (MBD) >500 (75)Delta P. (Ratio) >5

GVF% >90%

{l} >95% with liquid recirculation{2} Total volume at Feed Conditions{3} Limited by separation in dummy well and ESP capacity

Lor M H or MLor M H or M

M LM H

MHH

H

HH

L

<lOO (15)<2

<50%

N/A{5}H

M

500- 1005 to 2

90% - 50%

BP Multiphase Design Manual Section I7. Pressure Boosting

17.2 Pressure Boosting - Methodology of Selection ‘BOAST’

The BOAST (Boosting Options Analysis Software Tool) program, which has been developedwithin BPX for Foinaven, can be used to analyse the production profile achievable by the appli-cation of pressure boosting techniques for a particular case. The model takes account of threemain areas:

ReservoirCurves of reservoir pressure, water cut, GOR, and PI, each expressed as a function of pro-duced reserves, are loaded into the program. Using this data, the program can calculate theentry conditions into the well-bore throughout field life as the reserves are produced. In addi-tion the program can restrict the maximum production into an individual well; this later restric-tion is to incorporate reservoir drawdown restrictions such as imposed by gas and/or waterconing and sand production.

Production SystemDetails of the well-bore, flow lines and risers (length, diameter, number etc.) are entered into theprogram. Using standard multiphase flow correlations, the back pressure of the system for arange of flow rates can be calculated. Also included in the software is a description of the vari-ous boosting techniques to be analysed (generalised pump curves, maximum flow rate, maxi-mum power demand, maximum pump gas void fraction etc.) Using input data from the reser-voir description, the software is able to calculate the maximum flow rate achievable for the vari-ous boosting options.

The boosting options available are as follows:

l Well gas liftl Electrical Submersible Pumpsl Wellhead Multiphase Pumpsl Sub-sea Separation (with single phase pressure boosting)l Riser gas lift

Host FacilityThe software takes into account the capacity constraints of the host facilities and is able to limitproduction rates from the field if oil, gas, water, gross fluids or power constraints are exceeded.

Families of production profiles can be generated for a given field, production systemconfiguration, and host facility constraints for a range of boosting options, and thus a structuredchoice can be made. Note that the assumptions concerning multiphase pump capacities andperformance is based on FRAMO information and will need to be broadened to include thegreater experience now available.

For further information on BOAST refer to Multiphase Production Systems, TransportationGroup, BPX. at Sunbury.


Section 17. Pressure Boostinn BP Multiphase Design Munual


LEFT BLANK


BP Mult@hase Design Manual Section I7. Pressure Boosting

17.3 Outline Review of Multiphase Pumping Systems

17.3.1 Introduction

Pumps may be of either rotodynamic or positive displacement type. The two types operate ondifferent principles and it is practical to present their performance curves differently.

Rotodynamic pumps operate by imparting velocity (kinetic energy) to the fluid and then deceler-ating it to convert the kinetic energy to pressure. Consequently at any particular flow rate agiven specific energy or head is imparted to the fluid. Thus the differential pressure generatedis proportional to fluid density. The governing variable is flow rate through the pump and thecharacteristic is represented as shown in Figure 1 a below. If differential pressure is to beshown then the characteristic depends on density as shown in Figure 1 b.

-

Head

\

Differentialpressure

Increasingfl density/

3Flow rate Flow rate

Figure 1 a and 1 b

Positive displacement pumps operate by trapping a pocket of fluid at the pump suction,carrying it through the pump and forcing it out at the discharge. The differential pressureachievable is limited by the strength of the pump to withstand pressure and the effect of thedifferential pressure on the internal leakage. The characteristic of such a pump is conventionallyshown as in Figure 2a.

Flowrate

Differentialpressure

Differential pressure Flow rate

Figure 2a and 2b

If the axes are interchanged to achieve a representation similar to that for rotodynamic pumpsthen a performance characteristic similar to that in Figure 2b is produced. Comparison withFigure 1 b shows the difference in characteristic between the two types of pump.


Section 17. Pressure Boosting BP Multiphase Design Manual

17.3.2 Application to Multiphase Duties

If the two pump types are applied to Multiphase duties their characteristics remain unchangedbut the effect of the presence of the gas phase must be considered.

(a) Rotodynamic Principle

If the rotodynamic pump is handling a mixture in which the two phases are well mixed it can beregarded as handling a fluid of density and compressibility between those of liquid and gas.

In order to calculate the head it is necessary to consider the work done on the compressiblegas and the incompressible liquid separately. Even in cases of high GVF the mass of liquidpresent is much larger than the mass of gas and isothermal compression may be assumed.

This leads to:

(1-cx)(P,- P,) + aP,Ln (P,/P,)H=

PM

where:

pM = Mixture density = (1 -ol)pL + ap,

H = Heada = Gas Volume Fraction (GVF)P = Absolute pressurep = Density

Suffices: 1 = Suction2 = DeliveryG = GasL = LiquidM = Mixture

Analysis of published data shows that for a wide range of GVF a single curve represents thepump performance at a given speed.

Conversion of performance from one speed to another can be done using the normal affinitylaws:

l Capacity is proportional to speedl Head is proportional to speed squared



The effect of speed change is shown in Figure 3a below.

Increasing

Head

Flow rate

Figure 3a

(b) Positive Displacement Principle

Positive displacement machines trap a fixed volume of fluid at the suction and in general itmakes little difference to the capacity whether the fluid is liquid or gas. As speed changesinternal leakage remains constant while capacity varies linearly with speed. See Figure 4.

Twin screw machines tend to show increasing volumetric efficiency as GVF rises until at highGVF, typically more than 85%, volumetric efficiency starts to drop. This is thought to occurwhen there is no longer sufficient liquid present to fill the internal clearances and provide a seal.The lower density gas can leak through these clearances much faster than the liquid. The valueof GVF at which volumetric efficiency is a maximum depends upon the number of pitches onthe screw.

A typical curve is shown in Figure 5.

Increasing

Head

Flow rate

Figure 4

Volumeefficiency

0 100%GVF

Figure 5


Section I7. Pressure Boosting BP Multiphase LIesi@ Manual

Positive Displacement pumps tend to be of lower capacity than Rotodynamic but have agreater differential pressure capability and are less affected by changes in GVF, SuctionPressure and Differential Pressure. They are tolerant of viscous fluids but subject to wear bysand laden fluids.

Rotodynamic pumps have larger capacity but are much more affected by changes in GVF,Suction Pressure and Differential Pressure. Because head is a function of GVF, DifferentialPressure and Pressure Ratio, plots of differential pressure performance can only be preparedfor a given set of conditions. Viscous fluids degrade rotodynamic pump performance.

Because they do not depend on maintaining close clearances to maintain volumetric efficiencyRotodynamic pumps should be tolerant of sand laden fluids. However this has yet to bedemonstrated and experience with other pumps running at similar speeds suggests thaterosion may be a problem. The use of very hard materials may prove necessary.

(d) Rotodynamic Pumps

Helico-AxialThe leading rotodynamic multiphase pump is the helico-axial style of pump. The IFP designed‘Poseidon’ pump and its derivatives have been the most successful. Extensive testing andproduction operation has been carried out on versions of this pump supplied by FRAMO (FrankMohn) and Sulzer. It is most appropriate where high volume throughput is combined with highGVF at low pressure ratios, or where the GVF is lower and higher pressure ratios are required.The pump consisted of three basic modules:

(a) The Drive unit; this is generally an electric motor driving through a speed increasinggearbox; suitable electric motors for subsea application are under development (1 Q96) bySulzer, the Nautilus Project, and FFIAMO, their ELSMUBS Project. The SMUBS pump hasa hydraulylic turbine drive, see Section 17.4.1.

(b) The Pump consists of a multistage assembly of impellers which have axial flow and highsolidity. Shaft speed is high, typically in the range 4000-6000 rpm.

(c) The Homogeniser: it was found that pump performance was significantly improved if gasand liquid were well mixed; a homogeniser has been developed and is fitted immediatelybefore the pump suction.

This form of pump is establishing a good track record of successful operation, see 17.4.1, andis considered to be ‘commercial’, at least in the ‘medium’ size range, say up to 500m3/h.

Side ChannelSide channel pumps have high head/low flow characteristic combined with good gas handlingability. However they are of low efficiency and are used in low flow applications where theparticular characteristics are valuable and the power penalty can be accepted. They alsodepend upon close internal clearances to control internal leakage and are therefore unsuitablefor liquids containing abrasives, such as sand. It is expected that larger size pumps couldtolerate larger clearances while maintaining acceptable efficiency, but no suitable pumps arepresently available, and development effort has ceased. No commercial pump is in prospect.



Contra-RotatingTwo forms of multistage pump having contra-rotating impellers have been considered. Onehas stages arranged radially, similar to the Ljungstrom steam turbine, and the other type hasstages arranged axially. The former machine is designated as Contra-Rotating DiskCompressor (CRD). Performance of a test machine was said to have been good on gas butpoor when liquid was injected and development is either on hold or has been abandoned.The Contra-Rotating Axial Flow Pump (CRA) has been built by Frank Mohn and tested by Shellat their De Lier field for some 2 months, apparently successfully, although development iscontinuing. In both cases the purpose of employing contra-rotation is to give very high relativevelocity between the two sets of blading so increasing the head rise per stage and minimisingthe number of stages required. Suitable application might be low/medium total volume, withhigh GVF and low/medium pressure ratio boosting.

CentrifugalConventional centrifugal pumps lose performance at GVF above about 10%. A modifiedimpeller was developed which allowed effective operation up to about 50%. The modificationinvolves the provision of holes in the shroud and gaps in the vanes which allow liquid to sweepgas away from the positions where it tends to accumulate.

Axial FlowSome work on axial flow impellers showed promising results in the mid 1980s and further workhas been concentrated on axial flow impellers having a high hub/tip ratio in order to minimisecentrifugal separation. The impeller has been described as resembling an inducer whichsuggests it is broadly similar to the Poseidon (Helico-Axial) impeller. Testing at Texaco’sHumble site is being progressed; claimed performance/efficiency figures should be viewedwith suspicion since the basis of calculation has not been made clear.

Pumps intended for down-hole application with gas handling capability, say in the GVF range30-60 % have been designed; the much higher suction pressure is responsible for the lowerGVF and improves pump performance by increasing gas density. It is probable that the rangeof application is narrow, but where boosting is required down-hole at low, but significant,GVF then modified centrifugal pumping can be considered. Texaco are proposing to usehydraulically driven centrifugal pumps in the horizontal wells of the Captain Field (low GVF,heavy oil).

7.3.4. Positive Displacement

Twin ScrewTwin screw pumps were among the first to be applied for pumping multiphase mixtures. Theirpositive displacement action gives them an inherent gas handling capability. The use of purelyrotary motion and the absence of valves and sliding components are all advantages comparedto other types of positive displacement pump. In addition this type of machine has the highestcapacity for a given machine size which is attractive for the flow rates involved in oil productionoperations. Twin screw pumps can generally handle up to about 95% GVF. Higher GVF can behandled by injecting liquid into the pump so that the liquid content within the pump is main-tained to provide a sealant and also provide cooling and lubrication. In many cases GVF is sohigh that it is more appropriate to regard the required machine as a wet gas compressor.Since the twin screw machine in its normal form does not have internal compression the powerconsumption is high compared to that of a compressor performing a similar duty. The differ-ence depends upon the GVF and the pressure ratio. Attempts have been made to incorporateinternal compression but so far the increased losses have outweighed the benefits.


Section 17. Pressure Boostinp BP Multiphase Design Manual

This type of application has led to several successful prototypes and a number of commercialand semi-commercial applications. This twin screw pump format is poorly suited to subseaapplication because of the problems of seals and timing gears, see Section 17.4.2. The twinscrew pump is best suited to low/medium throughput applications where the pressure ratio ordifferential pressure is high and GVF less than say 95%. Twin screw pumps in commercialoperation are noted in Section 17.4.2.

Internal Twin ScrewThis development is a variant of the Mono pump (or Moineau Screw Pump) in which the statoris allowed to rotate, and known as the ‘idler’. In principle this is a twin screw pump in whichone screw rotates inside the other. This has the advantage that all components run in bearingswhich reduces wear by reducing forces between rotor and idler; this allows much higherspeeds to be used than in a conventional Mono pump, thereby reducing pump size for a givencapacity and allowing the efficient use of hydraulic turbine drive. However during testing it wasfound that the elastomer of the idler could not stand the duty and further work has been aban-doned.

Reciprocating/Linear MotorA reciprocating pump driven by a linear electric motor has been tested for subsea application.The project is known as Multiphase Electric Pump Station (MEPS). The unit is large and heavyfor a relatively low capacity, 1 30m3/h. This machine is best suited to low flow rate, high pres-sure ratio, high differential pressure applications. Initial works testing was followed by testing atTexaco’s Humble facility and is said to have been successful.

DiaphragmThe machine is based on the pumping of hydraulic oil by a conventional pump. The oil is alter-nately pumped into and out of two pumping vessels where it acts on an elastomeric diaphragmto displace the multiphase mixture. Material selection was said to have been satisfactory butproblems were experienced in detecting diaphragm position at the end of stroke. These aresaid to have been overcome and the pump has been on long term test. For subsea applicationthe hydraulic pump would be immersed in the oil reservoir. The volumetric capacity is notexpected to be large.

HydroboosterThis system is based on a conventional centrifugal pump pumping separated liquid sequentiallyinto a number of pumping vessels to alternately draw in and displace the multiphase mixture.Thus each vessel acts as the cylinder of a reciprocating compressor in which the liquid acts asthe piston. A full scale prototype is said to have been successfully tested on hydrocarbongas/liquid mixture.

Double-action PistonDouble acting piston pumps would work but have too low capacity for the application. A largenumber of pumps would be required to operate in parallel.

Ram Slurry PumpThere is no reason why a slurry pump should perform any better than a conventional pistonpump except for its solids handling capability. The use of rams lowers capacity compared topiston pumps. Valves are generally leaky compared to those of conventional pumps.

-



17.3.5 Other Systems -Jet Pump

The jet pump has the potential of being a simple no-moving-part pressure booster in relativelylow flow applications. Because of its very low efficiency it does need a virtually ‘free’ source ofhigh pressure driving fluid, such as a high pressure well which would otherwise be choked.It has the advantage of simplicity but high fluid velocities render it susceptible to erosion by

sand laden fluids; however by careful design this may be minimised and the vulnerable partsmade easy to replace. The high pressure motive flow should be single phase, otherwise thealready poor efficiency is further degraded. The motive flow can be gas for a gas-pressureboosting but for the multiphase flow pressure boosting the motive fluid must be liquid-only forreasonable performance. In the WELLCOM example it is proposed that the liquid from a highpressure well is separated from the vapour fraction and used to boost the pressure of a lowpressure well; otherwise the power consumption is high for a given pumping duty.

17.3.6 Systems with Separation

Strictly speaking these are not ‘multiphase’ boosting systems but depend on the division of theflow into separate, if relatively impure, phases:

Kvaerner Booster Station (KBS)The Kvaerner Booster Station is a separation/compressor/pump system based on a compactvertical separator with centrifugal pump below to increase the liquid phase pressure and a wetgas compressor to increase the vapour phase pressure. Although pilot trials have beensatisfactorily completed and the potential is for a very flexible system of wide applicability,the complexity of the system does not bode well for its commercial success. Furtherdevelopment is currently (3Q-‘95) on-hold.

VASPSThe Vertical Annular Separation and Pumping System (VASPS) is a way round the problem ofmultiphase pumping by separating the mixture and using conventional liquid pumps to handlethe liquid while allowing the gas to free-flow under its remaining pressure. Only the liquid phaseis pressure boosted so strictly speaking it is not a ‘multiphase pump system’. The separatorconsists of a vertical length of conductor some 600 mm in diameter. Multiphase fluid is fed intothis conductor and forced to follow a helical path. This induces accelerations of 3-4 g. Thehigh acceleration combined with the short distance bubbles must travel through the liquid leadsto good separation. The operating principle is thus that of the cyclone separator. Gas risesnaturally to the top of the conductor and flows away under its own pressure. Liquid collects atthe bottom and is pumped away by a conventional ESP installed at the bottom of a tube withinthe conductor.

-


Section 17. Pressure Boosting


LEFT BLANK

-



17.4 ‘Known’ Performance Track Record

Of the many proposed multiphase pumping systems only two have any significant track recordof industrial, (semi-) commercial operation, ie. the Helico-Axial and the Twin Screw type pumps.Also the Jet Pump principle has been in use in the WELLCOM system.

17.4.1 Rotodynamic - He&o-axial Pumps

The Poseidon Project was a French (IFP/Total/Statoil) led project to develop remote subseaproduction systems by identifying gaps in technology and developing the necessary techniquesand equipment. Subsea multiphase pumping was always regarded as an essential part of sucha system. The project itself was closed out in ca. 1990. A pump had been tested successfully(ca. 3000 running hours) at a land site in Tunisia. A second test in a Norwegian fjord in 1987-88 in 150 m water depth was also successful (ca. 4000 hrs).

Manufacture has been licensed to Sulzer Pumps and to FRAMO and several further successfuldevelopments have been carried out:

Pecorade Field (Elf)A pump (Model P 302) built by Sulzer was installed in the Pecorade Field in the Southwest ofFrance and has been running since June ‘94, with over 8000 hours by Feburary ‘96 The pumpis used in normal production operations but is also the subject of a test programme so that itexperiences a wider range of operating conditions than would otherwise be the case. Themaximum throughput has been ca. 365 m3/h, 55 MBD. There is an active proposal to continuethe operation of this pump on a commercial basis.

Gullfaks Field (Statoil)A pump built by FRAMO was installed on the Gullfaks A platform and started up in May 1994.It had accumulated over 5000 hours running by Mid ‘95, by which time the programme wasconsidered complete. Operation is said to be satisfactory except that seal leakage is consid-ered a bit high. Seals are manufactured by Burgmann and an in house (Burgmann) testprogramme is in progress aimed at improving performance.

Draugen Field (BP/Shell/Statoil)A SMUBS (Shell Multiphase Underwater Boosting System) pump has been installed on theDraugen Field in 270 m water depth. In the SMUBS Project, FRAMO, supported by Shell, havetaken the Guinard pump, which is very similar to the Poseidon Pump, and packaged it forsubsea installation. Pump internals are built into a tubular casing which can be installed fromthe surface into a vertical housing without diver intervention. Drive is by hydraulic turbine thusavoiding subsea electrical equipment and the problems of keeping well fluids away from themotor. There are technical and economic distance limits for power transmission. Injectionwater is being used as power fluid, and this SMUBS pump has now been running periodicallysince early November ‘94. All sub-sea operating procedures are said to have been verified andoperation is satisfactory. The use of the pump is said to have increased oil production by 39%;however the reasons for its periods of non-operation are not known for certain, although theyare believed to be due to platform production limits and the process requirement for maximumwater injection pressure.


Section I7. Pressure Boosting BP Multiphase Design Muntiul

17.42 Positive Displacement - Twin Screw Pumps

Twin screw pumps were among the first to be applied for pumping multiphase mixtures andseveral successful trials have been carried out, although the helico-axial pump is nowsurpassing it in terms of operational track-record.

Bokor B SarawakA Multiphase Systems MP 40 size pump (as on Forties Bravo) was tested on the Shell Bokor BPlatform, Off-shore Sarawak, at up to 97% GVF. Production operation was regarded assuccessful, after initial commissioning problems, with availability approximately 96%, and anestimated 16% increase in oil production.

Wehrblec, Germany and Housemartin, CanadaOn the Wehrblec Field in Germany a Bornemann Twin-Screw pump has been installed and asthe result the field life has been extended by ca. 5 years. On the Housemartin Field in Canada asimilar twin-screw pump has been installed to lower the well-head pressure and so avoidhydrate conditions.

Tinmar, TrinidadOn two wells in the Tinmar Field in off-shore Trinidad, operated by Mobil, Leistritz twin screwpumps (model L4) have been installed. This has resulted in the re-opening of one well and theup-rating of the other with a net increase in production and a pay-back time of less than amonth.

17.4.2 Jet Pump (WELLCOM)

The jet pump principle is in use as a multiphase pressure boosting system using the highpressure flow from one source, a well or group of wells, to boost the pressure from another,adjacent source, a well or group of wells.



17.5 Outline Design

17.5.1 Required Information

The most important parameters for the selection and sizing of multiphase pumps are:

Suction pressure P,Delivery Pressure P,Total Volumetric Flow at Suction Conditions Q,Gas Volume Fraction at Suction Conditions a

Because of its effect upon a and Q,, P, has a marked influence on the sizing of the pump andthe power requirement.

Other important considerations are the location, ie. subsea or surface, the power source andwhether abrasive solids are present.

[Note that the term Gas Liquid Ratio (GLR) is often used in connection with multiphase pumpsbut refers to volumes at suction conditions, as is GVF, rather than the accepted definition ofGLR with volumes referred to standard conditions; care must be taken to avoid this confusion.Note also that flow rates for pumps often use the units of BPD, which includes not only theliquid rate but also the gas, both at suction conditions].

Additional factors such as design temperature and pressure, and fluid composition/properties,will be relevant to detail design, pump material selection and motive power unit design. Theflow regime of the feed flow may dictate a need for the pre-conditioning of the feed, particularlyif sudden changes in GVF or periods of 100% GVF are possible. Furthermore the change induty, flow rate and suction/discharge conditions during operation may mean that no singlepump type, or even one form of pressure boosting, let alone a single pump size, will be idealthroughout field life.

The first stage is to calculate the maximum total flow at the suction conditions, (Q ,), the pres-sure ratio required (P,/P,) and the GVF, at suction conditions. An attempt can now be madeto select a suitable pump type from Table 2. However no pump type may be completelyappropriate and some accommodation may be necessary between what is ideally required andwhat is available.

17.5.2 mpe Selection

There is as yet insufficient information or operational experience to be definitive about the selec-tion of suitable multiphase pumping systems. The attached table, Table 2, shows the currentstatus of the systems showing some industrial potential, but some of these may fall-by-the-wayside and others be added to the list in the medium-term future.

Issue 2 - December 1994 17-

Section 17. Pressure Boosting BP Multiphase Design Munzd

GVF SuctionVolume

Press. ratioHigh/Med./Low

Multiphase pump type Other factorsPower Sand ? ‘Efficiency’ Sub-sea ?

High V.High1 oo-94% >1200m3/h

High1 200.750m3/h

H/M/L - Separation - KBS etc. E No* Good Yes

YesYesYes

?Yes

?Yes

?

YesYesYes

?Yes

?

Yes

YesYes

?Yes

?Yes

High x5 Separation - KBS etc. E No* GoodMedium 5-2 Helico-axial E or H Yes Moderate

Low <2 Helico-axial Eor H Yes Mod.

Medium High >5 Twin screw(tiq. recycle may be needed) E No Mod.750-250m3/h Med ium 5-2 Helico-axial or E or H yes Mod.

Twin screw(tiq. recycle may be needed) E No Mod.Low <2 Helico-axial or Eor H Yes Mod.

CRA EorH No Mod.

Low High x5 Piston/MEPS E No Good<250m3/h Med ium 5-2 Helico-axial or E or H Yes Mod.

Piston/MEPS or E No GoodCRA EorH No Mod.

Low <2 Helico-axial or E or H Yes Mod.CRA EorH No Mod.

- _ _ _ _ ~--

Medium V.High H/M/L - Separation - KBS etc. E No* Good94.70%

High High <5 Separation - KBS etc. E No* GoodM/L <5 Helico-axial E or H Yes Mod.

Medium High >5 Twin screw E No Mod.Medium 5-2 Helico-axial or Eor H Yes Mod.

Twin screw E No Mod.Low <2 Helico-axial E or H Yes Mod.

Low High >5 Piston/MEPS or E No GoodTwin screw E No Mod.

Med ium 5-2 Helico-axial or E or H Yes Mod.Piston/MEPS E No Good

Low <2 Jet or H No PoorHelico-axial E or H Yes Mod.

-

H/M/L - Separation - KBS etc. E No* Good

High >5 Separation - KBS etc. E No’ GoodM/L <5 Helico-axial E or H Yes Mod.

High >5 Twin screw E No Mod.Medium 5-2 Helico-axial or E or H Yes Mod.

Twin screw E No Mod.Low <2 Helico-axial Eor H Yes Mod.

Low High >5 Twin screw or E No Mod.Piston/MEPS E No Good

Medium 5-2 Twin screw or E No Mod.Helico-axial E or H Yes Mod.

Low <2 Jet or H No PoorHelico-axial E or H Yes Mod.

H/M/L - Separation - KBS etc. E No* Good

High >5 Separation - KBS etc. E No* GoodM/L <5 Modified (multistage) centrifugal pump EorH Yes Mod.

M/L High <5 Twin screw E No Mod.Medium 5-2 Modified (multistage) centrifugal pump E or H Yes Mod.

Low <2 Jet or H No PoorModified (multistage) centrifugal pump Eor H Yes Mod.

Notes: Sand * Small amounts may be acceptable, but disposable may be problematic‘Efficiency’ Thrs is a reflection of the apparent energy efficiency to transport oil; it does not imply anything about the thermodynamic effiaency, see 17.1Power E = Electrical H = Hydrauk

Yes?

YesYesYesYes

Low70.50%

V.High

High

Yes

YesYes

?Yes

?Yes

?Yes

?YesYesYes

V.low<50%

V.High

High

Yes

YesYes

?YesYesYes

Table 2. Multiphase pump selection chart


BP Mzdtiphase Design Manual Section 17. Pressure Boosting

17.53 ‘Feasibility’ Design

The feasibility of using a rotodynamic pump can be checked as follows:

1 Calculate maximum total flow at suction conditions. Figure. 5 shows the range of pumpsizes available. The maximum size for a single pump is presently quoted (by Sulzer Pumps)as about 1 200m3/h, (180 MBD = MPPG. in Figure 5.); however a limit of 500m3/h(75 MBD = MPP3/4) is shown in Table1 . Above this flow multiple pumps in parallel arepresently recommended

200,000(1325)

180,000(1192)

160,000(1060)

0’80,000

(530)

P302 QPecorade

40,000(265)

1

(132) MPPl M P P 2 M P P 3 M P P 4 M P P 5 MPP6

tPump size

P302 QPecorade

Figure 5. Maximum volumetric flowrate at suction


Section 17. Pressure Boosting BP Multiphase Design Manual

2 Calculate the pressure ratio required (PJP,) and verify from Fig. 6 that it is achievable.

P, = absolute suction pressureP, = discharge pressure

-

-

:

\

4s;

\

-

-

-

-

-

f\

\

<

\

-

P, ParIkk 6.6.

10 10 n

\ 16*16*kk 26~26~

'u'u 440440

1 22 33 44 55 66 77 88 99 10 GLR

= cl/(1 -cI)

Figure 6. Maximum achievable pressure ratio P,/P,

-



3 Assuming that these steps have shown the use of a rotodynamic pump to be feasible thepower requirement can be estimated from Figure 7.

5050

4040

3030

P,P, (bar)(bar)

2020

P, = absolute suction pressureP, = absolute suction pressureP,P, = discharge pressure = discharge pressure II

0.2800.280

0.2700.270

0.2550.255

0.2400.240

0.2000.200

0.1600.160

O],),,,~,,,,,,,,,,,,)00 11 22 33 44 55 66 77 88 99 1010 1111 1212 1313 1414 1515 1616 1717 1818 GLRGLR

p2/p,p2/p, 2.5 2.5

0.60.6

0.50.5

0.40.4

0.30.3

0.20.2

0.10.1

0 10 1 2 3 4 5 6 72 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 8 9 10 11 12 13 14 15 16 17 18 GLRGLR

Figure 7. Power coefficients

Issue 2 Issue 2 -- December 1994 December 1994 2 12 1

Section 17. Pressure Boosting BP Multiphase Design Munual

Read:

B = f(P, ,GLR) from the upper graphE = f(P, / P, ,GLR) from the lower graph

Power = ExQ,xP,kw

B

Note in this equation:

Q, = Liquid flow rate, m3/h (QL = Q, (1 -a))P, = Suction Pressure, barGLR = Gas/Liquid Ratio = al (1-a)

Note that published data appears to show a marked improvement in the performance of roto-dynamic pumps as suction pressure increases. This is because such data is normallypresented as graphs of Differential Pressure vs. Flow Rate. As the suction pressure increasesthe differential pressure can increase while keeping the head constant.

The feasibility of using a twin screw pump can be checked as follows:

1 Calculate the total suction flow rate.Twin screw pumps are obtainable up to about 2000m3/h, (300,000 bpd)

2 Check that the pressure requirements are within the capability of the pump beingconsidered, ie. pressure rating of the casing, seals etc.

3 Estimate the power required from

Power = (P2- P,) x Q,kW

36 x 0.6

Note: -

1 In this equation pressures are measured in bars and flow rate in m3/hr.

2 The 0.6 factor assumes an efficiency of 60%

3 Twin screw pump vendors’ machines vary widely in size and pressure rating.It would be best to check possible applications with one or more vendors.

It is possible that the power transmission limitations may restrict the available pumpperformance and will almost certainly affect the pressure boosting economics.It is necessary to consult electrical and control engineers on this aspect.


BP Multiphase Design Munual Section 18. Slug-catcher Design

Section 18 Slug-catcher Design

18.1 Introduction

18.2 Selection and Design

18.3 Vessel vpe

18.4 Pipe-finger mpe

18.5 Bubble-catching

18.6 Operation and Control

;..


Section 18. Slug-catcher Design BP Multiphase Design Manual

-


LEFT BLANK


Section 18. Slug-catcher Design

18.1 Introduction

The transportation of hydrocarbon fluids down a pipeline of any significant length often resultsin separation of the vapour and liquid phases such that the fluids travel along the pipeline inpartially separated packages, known as ‘slugs’, which are predominantly liquid, and ‘bubbles’,which are predominantly vapour. This is usually known as ‘slug flow’, and generally occurs inlines with a high liquid content. Alternatively with a predominately vapour flow, liquids mayseparate and settle in the low points in the pipeline; these pools of liquid may subsequently beprojected forward as slugs. This can happen either during an increase in vapour flow rate,known as ‘bean-up’ slugs, or when the line is pigged or sphered for any reason, often toprevent the excessive build-up of liquids in the line, thus producing ‘pigged’ slugs.

In normal slug flow (see Section 3.3 and 3.4) regular or random slugs occur virtually continu-ously and although the size of an individual slug is not predictable the statistical size distributioncan be calculated. In the vapour flow-lines the sizes of the pigged slugs, and to a lesser extentthe bean-up slugs, can be calculated; in particular the size of the pigged slug may determinethe frequency of the pigging operation, see Section 4.4. Slugs produced by transient effectsare covered in Section 7.

The phenomenon of slugging has both mechanical and process implications. Mechanically theweight and momentum of the liquid slug will put additional strains on the pipe work supportstructures, in particular because of the rapid changes in forces when the slug/bubble bound-ary passes. The process problems are caused by the rapid changes in duty/rate required bythe down stream processing facilities. The solution to these problems is a slug-catcher; this islarge vessel placed at the end of a multiphase line to enhance the vapour/liquid separation andto catch and store the liquid slug so that it can be passed forward for processing at a moresteady rate.

Even if a multiphase pipeline running under its normal operating conditions does not suffer slugflow, it should be remembered that start-up, up-set, low-flow or shut-down conditions mayresult in the productions of slugs, which need to be addressed at the design stage.

Note that, correctly designed and operated, a slug-catcher system will yield a nearly steadyforward feed of liquid but that the vapour phase will normally still be relatively unsteady.However it is possible to design the vessel with ‘over-pressure’ capacity to catch the gasbubbles, see Section 8.5.

Almost as important as the correct sizing of a slug-catcher is the operational control of thesystem; this is described in Sections 18.5 and 18.6.

-



-


LEFT BLANK

-.

4 - issue 2 - December 1994


18.2 Selection and Design

Once the occurrence of slug flow, or other unsteady flow phenomena requiring the use of aslug-catcher, has been identified the following procedure should be followed:

Step 1Estimate/Calculate Volume of the Slug, ie. the liquid volume of the slug; it can be assumed thatthe slug-catcher will carry out a basic vapour/liquid separation and although the vapour maycontain small amounts of liquid (say up to 10 USgal/MMSCFD or 0.001 kg liquid per kg vapour),it can be assumed that the liquid product is essentially gas-free, ie. contains no free vapourphase. It is initially assumed that no liquid/liquid (oil/water) separation takes place within theslug-catcher. The vapour stream can generally be sent forward to a scrubber for the removalof the remaining liquid before further processing, eg. dehydration, dew-point control and/orcompression.

Step 2Decide on the type of Slug-Catching Vessel required; this may either be a pipe-finger or avessel type of slug-catcher. See the attached Decision Tree Chart, Figure 1.

Issue 2 - December 1994 5-


very limited/costly

moderately Ilmitedlcostly. eg. fnshore

NO

f --zq ----4 L

Gas/ Condensate

-

-

Figure 1 Decision Tree Chart



Step 3The total liquid retention volume of the vessel has next to be calculated by adding tothe liquid slug size calculated (Step 1) additional volumes for control and operational safety:These will vary depending on the form of the down-stream processing of the liquid phases;the processing may either be continuous, at as near as possible to a steady rate, matchingthe pipeline average liquid rate, or it may be batch-wise, with periods of no liquid processingbetween the slug arrivals. The former would be typical of a black-oil pipeline; the later typical ofa wet-gas pipeline.

In a slug-catcher the slug size determines the volume between the High Liquid Level (HLL) andthe Low Liquid Level (LLL). The HLL is the level normally used as the design liquid level for thedetermination of the vapour handling capacity, defining the normal feed to the down-streamvapour handling equipment, and is the level at which an alarm will sound to warn the operatorof the high level for him to take action as appropriate, but no automatic executive action istaken. However a High-High Liquid Level (HHLL), a level at which executive action is taken toprotect down-stream equipment, needs to be determined. At this HHLL the down-streamequipment may not function to specification but the operation should not be hazardous.The volume between HLL and HHLL should be the maximum liquid flow multiplied by the timerequired to start-up the batch processing plus, say, one minute safety margin, or the time toclose the feed flow to the slug-catcher in a safe manner, whichever is the greater. If theprocessing is continuous, then it should be the maximum liquid flow rate multiplied by the timerequired to close the feed flow to the slug-catcher in a safe manner. The volume between theLLL, at which an alarm sounds but no executive action is taken, and the bottom of the vessel iscomposed of three parts: the ullage volume, the volume from there to the Low-Low LiquidLevel (LLLL), where executive action is taken to prevent gas breakthrough, and LLLL to LLL.The ullage volume is to choice but is a volume to hold sand, solids etc. and is assumed to takeno part in the retention or processing of the liquid phase(s). The LLLL to LLL volume dependson the form of the liquid down-stream processing; if the processing is batch-wise then it shouldbe the maximum liquid processing rate multiplied by the time taken to shut-down thatprocessing in a safe manner; if the processing is continuous then it should be the volumeassociated with the time taken to turn-down, in a safe manner, from the maximum to theminimum liquid processing rate, or it should be the maximum liquid processing rate multipliedby the time taken to shut-down that processing in a safe manner, which ever is the greater.The total liquid holding volume is now known.

Step 4The first pass at designing the slug-catcher can now be made. The first pass assumption isthat the vapour nominal velocity is 1 m/set, in the free part of the vessel, initially assumed to be50% of the cross-sectional area. The relevant vapour flow rate is the maximum flow, ie. thehighest vapour velocity in the bubble flow in between the slugs and at the minimum operatingpressure.

Notes on the Decision Tree Chart:

Oil/Water SeparationDo you require immediate oil/water separation within the slug-catcher vessel ? Note that this isgenerally to be avoided in that good oil/water separation requires the maintenance of steadylevels whereas the essence of slug-catching is that the level(s) are inevitably varying.

Real EstateIs the land or platform upon which the slug-catcher is to be installed of high value and verylimited in extent, eg. on an offshore platform, or is it of value and/or somewhat limited eg. inshallow water or an onshore location with significant land values, or of more-or-less unlimitedextent, eg. an onshore location with low land values.



Fluid TypeWhat is the general nature of the fluid ? A Wet Gas, with low liquid content in a predominantlyvapour flow; this will generally only produce liquid slugs during pigging or sphering, but mayalso do so during increases in flow rate (‘bean-up’ slugs). A Gas-Condensate, or very highGOR oil flow, will generally produce pigging or bean-up slugs and also irregular terrain inducedslugs. A Black Oil with medium or low GOR is likely to produce regular/random slugging at alltimes.

Design PressureWhat is the design pressure of the vessel ? Pipe-finger type slug-catchers can be maderelatively economically (and can be site-built) for high pressure application. High pressureapplications are say above 70 bar, Medium pressure between 70 and 14 bar and Lowpressure application below say 14 bar. Note that any requirement to damp out the vapourphase flow rate variations may dictate a higher operating and design pressure than mightotherwise be required.

Type of Slugging:If the slugs are only produced by pigging or flow increase (bean-up) their arrival can be antici-pated and their size can be to some extent controlled, either by pigging frequency changes orlimiting the size of flow rate changes. Thus the size of the slug can be limited by operationalconstraints and the slug-catcher size reduced. If the slugging is produced randomly by a slugflow regime then the size distribution can be predicted but the size of an individual slug can notbe controlled within that distribution.

In some border-line cases it may be necessary to carry out a preliminary sizing of both thevessel and pipe-finger type slug-catchers to see which will be more suitable.

Note: If it is necessary to effect oil/water separation in the slug-catcher then this should bereferred to XFE. XTP. Sunbury.

-

-



18.3 Vessel vpe

If a vessel-type slug-catcher has been selected then, from the flow area calculated above, thediameter of the vessel is given, initially assumed to be 50% full and assuming a length/diameterratio of say 51. The liquid volume can be then calculated. If the liquid volume is inadequate forthe known slug volume and the control functions as outlined in Step 3, then either the designlevel can be increased to a maximum of 65% and/or the vessel lengthened up to maximum ofsay IO:1 . If the liquid volume still dominates the design consideration should be given tolimiting the slug size or increasing the liquid processing rate; otherwise the vessel should bedesigned only on the basis of the liquid holding capacity. Conversely if the liquid volumeavailable is greater than needed the design liquid levels may be lowered to say 30% but it mustbe remembered that the differences between the various levels must be a minimum of 1 OOmmin order that the level instruments can function adequately; this may dictate a minimum liquiddepth. The vessel length/diameter ratio can be reduced to a minimum of 4:1, if the liquid depthis limiting.

The vessel design for slug-catching should be as simple and robust as possible in view of thehammering which internals are likely to suffer as a result of the liquid loads and slugs at highvelocities. The vapour/liquid separation performance can be checked by using the BP softwarePROSEPARATOR. Note that this should be used on the basis of a vessel with no internals, theliquid at a suitable elevated level, and the maximum vapour velocity.


Section 18. Slugxatcher Design BP Multiphase Design Manuul


LEFT BLANK



18.4 Pipe-finger Type

If a pipe-finger slug-catcher has been selected the number of standard large diameter tubes tohandle the vapour flow, as described above, is calculated eg. 36”, 42” and 48” tubes, (largersizes may be used, but they are not generally readily available). A smaller number of large boretubes is generally the more economic solution. The number of tubes is then doubled (to allowfor the tubes being 50% full). However the result must be a power of 2, ie. 2, 4, 8, etc. so thatappropriate dead-end tees may be used to sub-divide the flow as uniformly as possible.See Section 11 for more information on dividing multiphase flows. A typical Pipe-fingerSlug-catcher is shown in Figure 2.

StorageLiquid

Manifold

Gas Out

uidut

’ Manifold

Figure 2 Pipe-finger Type Slug-catcher

The initial length of the tubes should be taken as a minimum of 15 times the diameter. If theresult is a reasonable number then an initial design can be produced by linking the ends of thetubes with a transverse tube, usually of the same pipe-size, to act as a collecting vessel; thevolume of this is ignored for the first pass. The lower half of the main tubes is used for theliquid storage but if this is not adequate then additional tubing length may be added either tothe separation tubes or as an extension to the transverse tube. If drainage into lower tubes isrequired to hold the liquid volume then large bore, free drainage facilities into the transversecollection tube and the storage tubes must be provided.

If the number of tubes (calculated for the vapour handling duty) of the maximum size availableis very large, ie. more than say 8, then the assumed design liquid level in the tubes can bereduced to say 25% of the cross-sectional area, and the number of tubes re-calculated. Fullbore drainage from the main tubes to the liquid storage tubes must then be provided to drainthe liquid slug from the vapour separation tubes as quickly as possible.

The design of the vapour off-take pipes should be large enough to give a low initial verticalvelocity to maximise droplet separation. The preferable size is an equal Tee, ie. the same sizeas the separation / storage tubs, but should be no smaller than 0.7D, ie. area equal to the flowarea of the separation / storage tubes at 50% full.



Slug-catcher pipe-finger design should be kept relatively simple. One of the advantages of thepipe-finger type is that it is classified as ‘pipe-work’ for the purposes of design and inspection,with less onerous requirements; however this may be lost and the system re-classified as avessel if the design comes too complex, particularly if internals are added.

A possible hybrid design can be used with a ‘vessel’ designed for the vapour/liquid separationand pipe-work as the storage medium. Although more compact than the pipe-finger type andperhaps with a higher, more predictable separation performance than the pure pipe-finger type,it does not enjoy the full advantages of either the pipe-finger or vessel types. See Figure 3.

Gas BalanceStorageTubes

LiquidManifold

Gas Out

’ Feed

Figure 3 Vessel Type Slug-catcher

Where a slug-catcher is fed from a manifold or collection header, gathering the well-fluids froma number of wells and/or flow-lines, then care should be taken to arrange the pipe work suchthat the individual flow-lines are hydraulically disconnected from each other and from liquidsurges in the slug-catcher. This can be achieved by connecting the flow lines into the top ofthe header so that there is no direct liquid communication between the individual flow-lines andthe slug-catcher.


-


18.5 Bubble-catching

When slugging flow occurs the flow consists of trains of liquid slugs and vapour bubbles. Whilethe slugs of liquid can be caught by level variation in the slug-catcher vessel there is no suchvolumetric ‘catch’ of the vapour phase bubble. The best that can be achieved is damping ofthe vapour flow by pressure variation in the vessel, but the required additional pressure rating ofthe vessel can be expensive to provide.

In its simplest form smoothing the forward flow of vapour can be achieved by flow control of thevapour stream, rather than using the vapour release rate to determine the pressure within theslug-catcher. However there will be an upper limit to the pressure variation allowed within theslug-catcher according to its design, so that the flow controller set point must be re-set or over-ridden by the measured pressure.

Allowing the pressure in the slug-catcher has implictions on the other parts of the system whichmay be undesirable: a higher pressure will cause an increase in feed flow back-pressure andmight ultimately cause production restrictions. Changing the pressure has an effect on thequality (RVP, Bubble Point, etc.) of the liquid phase, which in some cases may cause otherprocessing difficulties. The extent to which bubble-catching by pressure variation might beuseful or detrimental can best be determined by carrying out a dynamic analysis on the slug-catcher operation, see below.

Issue 2 - December 1994- 13


.-


LEFT BLANK

-


--

-


18.6 Operation and Control

The control of a slug-catcher is fundamentally very simple but in practice rather more complexdue to unsteady nature of the operating parameters.

The operational principle is to carry out a rudimentary vapour/liquid separation and toaccumulate the liquid within the vessel to allow a steady forward flow of the liquid phase to thedown-stream process. This means that the liquid exit flow should be by flow control, ratherthan level control; this flow control will need to be set at the time-average liquid make, whichcan only be estimated from the historical records and or the choke positions/well test data.The liquid level will need to be monitored to avoid over-filling (carryover) or emptying (gasblow-by) and may be used as re-set value for the flow control if, but only if, the level movesoutside the acceptable range. A slug-catcher does not by its very nature have a level controlsystem - the level is variable or it is not acting as a slug-catcher. A historical plot of liquid leveland forward flow against time, possibly with a computed figure of inflow, calculated from chokepositions and well-test data (in the same units and at the same conditions as the measuredflow), would be useful aid to operator control.

The pressure in the slug-catcher can be controlled by regulation of the vapour off-take rate,in the same way as a standard V/L separator. However if the slug-catcher is also required tosmooth out the vapour flow this will not be adequate. The vapour exit flow should then also beflow controlled, reset by the vessel pressure.


Section 18. Slug-catcher Design BP Multiphase Design Munual


LEFT BLANK

-


Section 19 Sandwashing

Design of Vessel De-sanding Systems

19.1 Introduction

19.2 Method Outline

19.3 Design Steps

19.4 Control and Monitoring

Appendix


Section 19. Sandwashing BP Multiphase Design Manual


LEFT BLANK

2 issue 3 - January 1998

19.1 Introduction

Solids are often co-produced with hydrocarbon fluids, such as sand, from unconsolidated sandreservoirs such as Forties, and Proppant, which is injected into the formation in the SouthernNorth Sea Gas Fields to keep open fissures. The solids are often associated with increasedwater production. This section describes a design method for a de-sanding system for horizon-tal three phase separation vessels while on-line, i.e. without emptying the vessel or even stop-ping oil production.

It is based on work carried out at Sheffield University, Department of Mechanical & ProcessEngineering, and Flow Systems Design Ltd. (J R Tippet& & G H Priestman), and the BP soft-ware subsequently written based on that technology. The concept is to inject wash-water to acreate a given momentum flux through a number of flat fan-jets carefully distributed through thelower part of the vessel, subdivided as necessary, followed by gravity drainage through a smallnumber of sand-wash drains.

-


Section I9 Sandwashing BP Multiphase Design Munual

-


LEFT BLANK


-

19.2 Method Outline

Basic Information

Before the design of a de-sanding system can be attempted certain basic information isrequired concerning the vessel to be sand-washed, the solids to be removed and the wateravailable to carry out the process. These data are listed in Table 1. The sand-wash method isbased on the use of high pressure water injected through 45” fan flat jet nozzles (“jets”) of thetype supplied by Lechler Ltd. of Sheffield, or equivalent, see Table 2.

Sand-Wash Sections and Tiers

Unless the vessel is small i.e. less than ca. 2m diameter with an L/D ratio of less than 3:1, thenthe vessel will need to be subdivided into sand-wash sections and will need more than one tierof sand-wash jets each side. The general arrangement is usually dictated by the design/size ofthe vessel, particularly the presence of weirs or baffles in the liquid part of the vessel, althoughin some cases it may be advisable to reconsider the choices first made in the light of laterresults.

FF Equation

The key tool in the design of the sand-wash system is the “Fluidisation Factor” or FF equation.Having determined the length of the sand-wash section and the number of tiers, a preliminaryfigure for the jet spacing and number of jets can be determined. Using the FF Equation, a jetsize is selected which determines the jet pressure drop required to give a Fluidisation Factor(FF). The jet size and number and the pressure drop give the total wash-water flow. By select-ing different jet sizes and numbers/spacing, the balance between jet pressure and water flowrate can by changed as required. However, if there is a fundamental mismatch it may be neces-sary to change the number of tiers or sand-wash section length.

Ancillary Calculations

Having determined the wash-water flow rate for a section, the header, wash-pipes and thesand drain can be calculated. If the quantity of sand in the feed fluids is known then an esti-mate of the required interval of de-sanding can be made. This may influence decisions aboutthe degree of automation in the control system.

The above functions are partially automated in an Excel Program SANDWASH.XLS,

issue 3 - January 1998 5

Section I9. Sandwashing


LEFT BLANK


Section 19. Sandwashing

19.3 Design Steps

Wash Sections

The length of a wash section, that is the length of the vessel serviced by a set of wash pipesand a single drain nozzle, can be up to three times the vessel diameter; this assumes that therequired volume and pressure of wash-water are available, that the drain can be placed in thecentre of the section and that the floor of the vessel is un-obstructed. One and a half vesseldiameters is the maximum distance of the end of a wash-section from the drain, hence themaximum wash section length of three diameters.

This generally will mean that no more than two or possibly three sand-wash sections, each withwash-pipes and drain nozzle, are required in a standard horizontal separator, unless processinternals dictate otherwise, or the wash-water feed rate or sand-wash drainage rate is veryrestricted. The existence of weirs or oil/water baffles for process reasons may already dividethe vessel into sections and it is generally advisable to make use of these sub-divisions.

The end of each section, unless bounded by an existing feature such as the vessel end orprocess baffle, is bounded by a solid segmental baffle, or sand-dam, in the bottom of the ves-sel, about 0.085D high. The purpose of the sand-dam is to restrict the flow of sand out of thesection being de-sanded and so settling in a neighbouring in-active section. The height of thissand-dam is not critical and a lower baffle may be necessary for other process reasons, suchas oil/water interface level, but a minimum height of 150mm is recommended, (see Fig.1 (a)).

The number of wash-sections is determined by inspection; the longest section is the most criti-cal and should be designed first. It is this section which will need the largest water flow and/orjetting pressure and will thus determine the size of the wash-water supply pumps and the efflu-ent handling system.

Tiers of Wash-Pipes

Depending on the internal diameter of the vessel, more than one tier of wash-pipes and jetsmay be needed. Unsupported sand will settle in the vessel on internals and on the walls up tothe point where the angle of the wall is the angle of repose, (see Fig.1 (b)). It is assumed thatthe initial depth of sand is not excessive, say less than 10% of the vessel volume. If the angle ofrepose is not known then the default figure of 34deg. should be used; this is typical ofsand-like solids.

The wash-pipes are fitted with the flat fan jets and should be positioned so that they sweep thebottom of the vessel from the 34” line along the side of the vessel down into the centre wherethe drain nozzle is located. The upper tier of jets should be on the longitudinal line at 34”. Thelateral “throw” of a jet is 1.67 times the longitudinal spacing; the maximum longitudinal spacingrecommended is 450mm so that the maximum lateral separation is 750mm. This correspond toa maximum vessel size for a single tier of jets of 2.5m. If the jet spacing is 300mm the largestdiameter vessel which can be de-sanded by one tier is 1.67m, for two tiers the maximum diam-eter is 4.5m. or 3.4m for 300mm jet spacing. If a large number of jets is required, giving a veryclose jet spacing along the wash-pipes, then it may be advisable to increase the number of tiersabove the minimum; the aim is to obtain a fairly even pattern of sand-wash jets.

issue 3 - Januarv 1998 7

Section 19. Sandwashing l3P Multiphase Design Mamml

Wash-Jet Nozzles

The sets of jets on either side of the centre line should be identical; therefore the number of jetsin each section must be divisible by two. The jets should be at an even spacing along thewash-pipe running the length of the sand-wash section, with half spacing at either end, (seeFig.3). If the numbers of jets in different tiers are not identical, i.e. because the number of jets inthe section is not divisible by four, (or six in the case of three tiers), then the smaller numbershould be in the upper tier, with full spacing between the jet and the end of the section, (seeFig.4). The maximum practical jet spacing is ca. 450mm, otherwise significant piles of sand willbe left between the jets. There is no minimum spacing but if the jets are closer than say150mm, larger jets, more widely spaced, or an additional wash-pipe tier might give a betterspatial arrangement of jets, and thus a more even momentum flux.

The 45” flat fan jets should be positioned so that the plane of the fan is parallel with the axis ofthe vessel and at 45” dip to the horizontal; the jets should be positioned 0.028D away from thevessel wall, (see Fig.1 (b)). This positioning is to avoid direct impingement and abrasion of thevessel wall or any lining. The fixing of the jets is critical -they must not be allowed to rotate sothat the fan impinges on the vessel wall, nor allowed to fall out, thus leaving a very large hole inthe wash-water pipe with a disastrous effect on the de-sanding efficiency and probable rapidand severe damage to the vessel walls.

The size of the jets is required in terms of the Effective Diameter, deff, i.e. that diameter whichwill gives the pressure drop/flow relationship with a discharge coefficient of unity. Lechler quotethe sizes in terms of the Equivalent Bore (EB); a list of EB v. deff for the relevant Lechler Jets isgiven in Table 2. The size of the jets selected should give the required FF considering thewash-water pressure and flow rate available, (see below). They should also be large enough toresist blockage by solid contamination in the wash-water and small enough not to suffer block-age by the ingress into the jets when submerged in the sand. A size (EB) range of 4mm to12mm is recommended.

F-Factor, Wash-Water Flow Rate and Pressure

The fundamental parameter for successful de-sanding is momentum flux; this is described bythe Fluidisation Factor “F” in the F equation. For sand a FF value of 0.001 is required; for prop-pant, preliminary indications are that a value of 0.0025 is required. It is probable that the valueof F is proportional to the size of the particles, but until this is confirmed an experimental valueof F must be determined for solids differing significantly from sand.

The value of FF assumes a depth of sand which is much larger than would normally be expect-ed, i.e. up to 10% of the vessel volume, (a depth of 15% D, see Fig.1 (a)).

Slightly lower values of F than those recommended will result in rather slower sand-wash rates,but significantly lower values will give incomplete solids removal, whatever the duration of thesand-washing. Values of FF greater than the recommended value will merely waste energy andwill not result in faster or more effective de-sanding, but will of course do little harm.

The FF equation is as follows:

FF = (P.Pj.n.deW2) / (Ls.D*2.(rhowhol).g)

where:


FF = Fluidisation FactorPj = Pressure Drop across Jetsn = Number of Nozzles in Sectiondeff = Effective Diameter of JetsLs = Length of SectionD = Internal Diameter of Vesselrhos = Density of Solidsrhol = Density of Wash LiquidQ = Wash-Water FlowPO = Operating PressurePd = Wash-Water Delivery Pressure

A preliminary number of nozzles suitable for the sand-wash section has already been deter-mined from the geometry of the vessel. Entering the value of FF, the number of nozzles, and afirst guess at the jet effective diameter in the FF equation gives the required pressure dropacross the nozzles; this also defines the wash-water flow rate (Q) by virtue of the known jet sizeand number:

Q = (n.deff2/4).n.(2.Pj/(rhos-rhol))*0.5

To obtain the necessary wash-water delivery pressure the operating pressure within the vesselplus an allowance for the sand-wash pipe-work must be added to the calculated jetting pres-sure drop. The vessel pressure when sand-washing is generally the normal operating pressure,and pressure drop loss might be say 10% of the jetting pressure; any higher and the mal-distri-bution between the jets becomes significant. Thus:

Pd=Po+l.l.Pj

The first estimate of the wash-water flow rate (Q) and pressure (Pd) may well not be satisfactoryand adjustments can be made to the size and number of jets. It should be remembered that ahigh wash-water rate also requires equivalent disposal facilities to treat the effluent from thevessel, whereas higher jetting pressures only require a higher energy in the wash-water feedsystem. The various courses of action are suggested below:

P & Q slightly too high:Reduce FT - this is only acceptable for a small reduction of FF and/or if the sand loading is lightand/or the sand is fine, as in a second stage separator.

P & Q too high:Reduce the size (length) of the wash section.

P & Q too low:Increase the size (length) of the wash section.

P too high and/or Q too low:Increase size and/or number of jets, check the space distribution.

P too low and/or Q too high:Reduce size and/or number of jets, check the space distribution.

Having determined the maximum jetting pressure and wash-water flow rate for this section of


Section 19. Sandwashing BP MultQhase Design Munual

the vessel, the wash-water delivery system, pump(s), supply pipe-work and internal headerscan be calculated. The sand drains can also be sized, (see below). It should be rememberedthat the sand-wash-water has to be disposed of, and the size of the system is determined bythe greatest flow. It is therefore advantageous to equalise the wash-water flows between thevarious sections as far as possible.

Drain Nozzles

The number of sand drain nozzles is given by the number of sand-wash sections. The nozzleshould be placed as near as possible to the centre of each sand-wash section but no furtherthan 1.5D from the end of the section.

The nozzles should be a minimum of 3”NB to avoid blockage by stones etc. which are found insome formations and should be fitted with sand caps; the design of these are shown in Fig. 2.

The size of the drain nozzles can be determined once the wash-water flow is known; the exitvelocity should not exceed 3m/sec. to limit the erosion. Larger nozzles may be required in somecases if the driving pressure differential is low.


19.4 Control and Monitoring

The de-sanding operation is very simple. The section to be sand-washed is selected and thewash-water is opened: flow and pressure of the wash-water flow are noted, see below. After acouple of minutes the appropriate sand drain is opened and the sand and wash-water drainedfrom the vessel either for a preset time based on experience and expected sand content or,preferably, until a monitor on the effluent flow indicates that the sand content of the effluent flowis low -the wash-water and sand drain are then closed.

Once the sand-wash system has been designed correctly, the amount of control of the systemfor efficient de-sanding operation is limited to monitoring. Measurements of the delivery pres-sure and flow rate of the wash-water can be used to monitor the status of system and givewarning of jet blockages or wear. This is very important as severe wear or even the loss of jetshas a disastrous effect on the de-sanding efficiency and can result in severe damage to thevessel walls.

The flow/pressure relationship for the system should be measured during commissioning andthis relationship can be used during subsequent operation to monitor the mechanical health ofthe system. During commisioning of the system with the pipework and jets in a good mechani-cal condition a series of measurements of wash-water flow rate against delivery pressure,preferably with the vessel filled and pressurized to standard operating condition, should betaken. With this data a graph of flow rate v. pressure can be drawn for comparison with subse-quent operations. A higher flow rate for a given pressure or a lower pressure for a given flowrate indicate either worn jets (gradual change) or even a missing nozzle (sudden change); thereverse symptoms indicate probable nozzle blockage(s). Depending on the severity of the devi-ation some remedial action may be required, probably involving vessel entry.

A measurement of the solids content of the effluent stream is also highly desirable in that it canbe used to determine the end of the useful sand-wash period and thus minimise the sand-washtime and costs and the disturbance to the oil/water interface in the vessel. While there is bulksolids in the vessel to be removed, the momentum of the jets is largely absorbed by the solidsand the disturbance to the oil/water interface is minimised and thus potential cross contamina-tion of the streams is minimised. Once the bulk sand has been removed, the concentration inthe effluent flow drops markedly to the sand content of the “stirred-up” water. An ICI TracercoDensitometer fitted on the effluent pipe-work has been shown to give this indication at theexperimental facility, but has yet to be proven industrially.

Issue 3 - January 1998 I I

Section 19. Sandwashing

Input Data Required.Non-essential but useful information is shown {...)

Vessel Dimensions:Diameter (ID)Length (T/T)Position of Internals in Liquid Phase{Length of sand-wash sections

Operation:Operating Pressure

Sand or similar solid:{Sand Feed RateAngle of ReposeDensity of solid

Wash Water:Restrictions (if any) on wash-waterflow rate and delivery pressure.Density of wash water

Table 1. linput data

Lechler Fiat Fan Jet Nozzles 45”

LechlerCode Nos.

Equivalent Effective DiameterBore “EB“ deff

0-t-W (mm)

670643 2.50 2.06670723 3.00 2.59670673 3.50 2.91670803 4.00 3.26670843 4.50 3.64670883 5.00 4.12670923 5.50 4.61670963 6.00 5.15670993 6.90 5.64671043 8.00 6.5267 1083 9.00 7.28671123 10.00 8.18671163 11.00 9.21671203 12.00 10.30

= D (m)= L (m)

= Ls (m)}

= P (bara)

= S (m3/sec.)}= a (if known, default =34”)= rhos (default for sand = 2600 kg/m3)

= rhol (default = 1050 kg/m3

Table 2. Lechler jet data

12 k.sue 3 - Januaw 1998

References

“Fundamentals and Design of Sand-wash Systems for Separator Vessels, Final Report,December 1990”, by J.R. Tippet& & G.H. Priestman, Flow Systems Design Ltd. and SheffieldUniversity, Department of Mechanical & Process Engineering.

Issue3 Issue3 -- January1998 January1998 13

Section 19. Sandwashing BP Multiphuse Design Manual


LEFT BLANK

-


Sand DamTo partition sand-wash sections

Sand-wash Jets2 tiers

Max. sand fill level0.15xD --

(for F = 0.001)

Fig. 1 (a)_-_-------w-v

\\\ \\ \ \ \ \ \\\

0.028D jet / inner wall separationI

pipes

Fig. 1 (b)

Sectio

n 19. S

an

dw

ash

ing

l3P M

ultiphase L)esig?l Manual

Fig

ure 2. S

and

drain

s.

-

I6Issu

e 3 - Janu

ary 1998

I

Fig

ure 3. S

and

-wash

jet spacin

g (1).

IIIiIIIIIIIIi1IIIiIIiIIIIiIII

Issue 3 - Jan

uary 1998

17

Sand-wash section

Wash-water i

feed header i

i

i‘45IIIiI

I Centre linei of vessel

-----------------------------------1.--___--

i

I-S&I

IIII

II

- - - - - -f

i

M a x = i

---------II

I I II II I ;I,+ s/2 -44 S d I

II II II I i

Arrangement of sand-wash pipes/jets (Number of jets not divisible by four)Jet spacing = S

BP Multiphase Des&11 Mailul Section 20. Deepwater multiphase flow

Section 20 Multiphase flow throughdeepwater systems

20.1

20.2

20.2.1

20.2.2

20.2.3

20.3

20.3.1

20.3.2

Introduction

Key technical issues

Energy

Integrity

Delivery

Deepwater operating limitations and solutions

Critical factors for deepwater hydraulic design

Assessment of limitations

References


Section 20. Deepwater multiphase flow BP Multiphase Design Munual


LEFT BLANK


-

Section 20. Deepwater multiphase flow

20.1 Introduction

A large proportion of BP’s future oil production is likely to be provided by deepwater offshoredevelopments. A DeepStar study estimated that 95% of presently unexplored offshore acreageis in water depths of 3,000 ft or more. The main areas of activity at present are in the Gulf ofMexico (GOM) where BP has a 100% stake in 267 licence blocks in water depths of over 1,000ft and is presently assessing developments in over 6,000 ft of water with current leasesextending to 9,250 ft. West of Shetland the water depth is not as great (around 1,600 ft atpresent with potential developments at depths up to 4,000 ft). However the environmentalconditions are very severe and have had a large impact on the design of the present Foinavenand future Schiehallion projects, and potentially for the latest discovery Suilven. The other maindeepwater development areas are offshore Norway, where BP is drilling a well in the Voringbasin at a water depth of 4,200 ft, and West Africa (600-7,000 ft) with present studies of the4,500 ft WD Girassol development. Although these are the main deepwater provinces for BPthere are also other potential future prospects in the Black Sea, Caspian Sea, and offshoreIndonesia, China and Vietnam. Each of these areas have their own unique combination ofconditions relating to the available oil and gas infrastructure, environmental conditions, andpolitical climate etc.. but share in common the deepwater nature of the province. Figure 1 illus-trates the environmental conditions characteristic of these areas of activity.

One of the critical issues is whether a tie-back to shallower water depths is practical as this mayeliminate the need for ultra-deepwater surface piercing structures. Hence we need to investi-gate the hydraulic limitations on the step-out distance, but how far do we need to go ? Atpresent the Troika subsea development requires 14 mile long flowlines from a seabed templatein 2,670 ft of water to the existing Bullwinkle platform in 1,350 ft of water. One of the controllingfeatures of the design is to maintain a high level of insulation on the flowlines in order to providehigh flowing temperatures and to avoid hydrate formation. The GOM Neptune discovery is in6,000-7,500 ft of water and could conceivably involve tie-back distances in the range 25 to 30miles. However this may not represent the ultimate challenge.

Drilling has already taken place in the GOM at water depths of 7,800 ft (BAJA), and the deepestlease in the April 1996 GOM Central lease sale was in 9,289 ft of water. Hence a realistic designchallenge is 10,000 it water depth. The Shell ‘Mensa’ development is in 5,300 ft of water andinvolves a 68 mile, 12” tieback to a host platform. Looking at the existing GOM infrastructureand the potential future plays indicates that a tie-back distance of 100 miles is a reasonabletarget. Hence we need to understand the limitations of pushing the design operating envelopeto water depths of 10,000 ft with tie-back distances of 100 miles.

If we can design for such long tie-backs this may eliminate the need for ultra-deepwater surfacepiercing structures. However it is likely that conventional fixed platform designs will be uneco-nomic and impractical at such water depths and hence some floating structure will be required.Although there are a large array of potential host platform configurations ranging from floatingtankers, tension leg platforms, spars, compliant towers, and semi-subs etc.. they share incommon the need for riser systems to connect the subsea flowlines to the facilities. Theserisers can be rigid vertical conventional or hybrid designs, steel catenaries, or may beconstructed of flexible pipe in a variety of configurations including catenary, ‘S’ and potentiallydouble ‘S’. As the water depth increases the pressure drop over the riser increases and has alarger impact on the hydraulic characteristics of the system. However there may be severe limi-tations in the ability of present design methods to accurately predict the hydraulic performanceof such deepwater riser systems, and there is no field scale data on which to test and evaluatethe performance of the models. The present limiting size of flexible risers is around 12”,although at increased water depths sizes of up to 20” may be required. This large diameter

issue 3 - Januarv 1998 3

Section 20. Deepwater multiphase flow BP Multiphase Des&n A4anual

brings into question the accuracy of the hydraulic models which have historically been validatedon relatively small bore well tubing.

This section of the Multiphase Design Manual is based on the results of a study into thehydraulic design issues relevant to deepwater developments which was undertaken in 1996under the sponsorship of the Deepwater MTL (Reference 1).



.-

20.2 Key ted-mid issues

The key technical issues can be summarised under the headings of energy, integrity anddelivery:

20.2.1 Energy

‘The provision of sufficient pressure to transport the required flowrates of fluids from the reser-voir to the processing facilities.’

The system pressure drop increases with water depth and step-out distance as illustrated inFigures 2 and 3. This may be a limiting factor if the natural well head flowing pressure is insuffi-cient. Most of the pressure drop is expected over the deepwater riser, although the proportionof the total pressure drop over the flowline will increase with the tie-back distance. A trueassessment of design limits requires accurate pressure loss prediction methods for which thereare two main concerns. Firstly, the shape of the flexible and steel risers utilised in deepwater isexpected to influence the hydraulic characteristics and there is a need for design tools andguidance on the effect of the different riser shapes, to enable accurate predictions to be madeand to assess the issues and limitations pertaining to the various configurations. One potentialproblem lies in the basis of the steady state multiphase design programs, such as Multiflo,which describe pipelines and risers by a number of discrete segments to capture the featuresof the topography. This may require many sections for a true representation of a catenary or ‘S’shaped riser. However each segment is treated in isolation and hence the flow regime andholdup can change significantly at the segment boundaries. In practice the conditions in onesegment can influence the downstream conditions, for example, slug flow generated in theflowline may persist in a riser, even though the riser is predicted to operate in another flowregime.

The second potential inaccuracy is in the fact that the hydraulic models developed for verticaland highly inclined flows have historically been developed from relatively small bore well tubingdata and have not been validated at the increased diameters that may be required for deep-water developments.

In deepwater developments the reservoir can fail to produce naturally at much higher pressuresthan conventional shallow water developments. Hence it is possible that a great benefit can beobtained by providing some kind of pressure boosting or pressure drop reduction. Pressureboosting can be obtained by using down hole ESP’s, multiphase pumps or subsea separationand boosting. Pressure drop reduction can be provided by using gas lift, drag reduction agents(DRA), using foaming agents, or removing the water at the wellhead. The optimisation of thesemethods requires that many factors be taken into consideration and can benefit greatly fromintegration with the reservoir performance, the flowline and riser hydraulics, and the processequipment performance. Steady state programs such as Multiflo are ideal for generating thehydraulic data provided that the accuracy is sufficient when applied to deepwater systems.Future system integrated models offer promise if the complexity does not hamper the simula-tion speed. This may require that a simple description is used for the reservoir performance ashas been adopted by the BOAST program developed for AFP. Figure 4 illustrates the mainenergy related issues.

Transient considerations may also play a part in the energy related issues in that unacceptableconditions may arise due to unstable operation under normal flowing conditions or during shut-downs, restarts and flowrate increases. For example sufficient pressure may not be available to

Issue3 Issue3 -- Januarv1998 Januarv1998

Section 20. Deepwater multiphasejlow BP Multiphase Design Mmual

start-up wells against the high back pressure generated by stationary fluids in the flowline andriser. Temporary riser gas lift may be a solution to unloading the riser and kicking-off the wellswhich may flow naturally after start-up.

20.2.1.1 System pressure drop components

The production capacity of a deepwater development is principally determined by the reservoirpressure available and the pressure drop over the entire system from the sand face to the facili-ties. If there is sufficient driving energy from the reservoir to overcome the total pressure drop tothe facilities the desired production capacity will be realised provided that the system integrity ismaintained and the facilities can handle the delivered fluid flowrates. Figure 5 illustrates an inte-grated approach to the overall system pressure drop and indicates the various componentsthat contribute to the system pressure drop.

(a) Reservoir

Starting at the reservoir the pressure is normally high initially and declines with time as thereservoir is depleted, although the pressure decline can in normal cases be moderated by pres-sure maintenance via water or gas injection for example. A reservoir that is normally pressuredhas an initial pressure approximately equal to the hydrostatic pressure for a column of waterequal to the depth of the reservoir. For example a 15,000 ft reservoir from the seabed locatedat a water depth of 10,000 ft would have an initial pressure of approximately 10,800 psi. Ifadequate pressure support is not provided the pressure could drop to the bubble point of say5000 psi. This will have a great influence on the energy available for transporting the fluids.

The range of reservoir characteristics and flowing conditions are great and it is beyond thescope of this manual to consider all options. This is best done when analysing specific develop-ments. The aim here is to illustrate the effect of various parameters pertinent to deepwatersystems and hence to illustrate the significance.

(b) Formation pressure drop

As the fluids are produced from the reservoir to the well bore a pressure drop is experiencedover the formation. The relationship between the formation pressure drop and the flowrate iscalled the inflow performance and in its simplest form is expressed as a linear relationshipbetween the oil flowrate and the draw-down, which may be typically IO to 100 bpd/psi. Hencea productivity index of 10 with an initial reservoir pressure of 10,800 psi would yield a bottomhole flowing pressure of 8,800 psi for a oil flowrate of 20 mbopd.

(c) Well tubing pressure loss

The next component of the system pressure drop is the tubing loss which in shallow watersystems is usually the biggest loss. However in deepwater the greater riser depth substantiallyincreases the significance of the riser pressure drop. In the early production life of an oil field theflow into the well bore may still be above the bubble point, but the flow usually becomes twophase due to the pressure drop in the well tubing. Hence we need two-phase flow models topredict the pressure drop. For gas wells retrograde condensation usually also gives rise to atwo-phase mixture.

Typical vertical two-phase flow patterns are illustrated in Section 1 where the flow may be

-



bubbly just above the bubble point in an oil well but may change to slug, churn, and finallyannular if there is sufficient gas evolution. The flow is generally symmetrical about the centre linedue to axial gravity effects.

There are three main components to the pressure drop across the well tubing and these aredue to friction, hydrostatic effects, and acceleration. At low flowrates the hydrostatic termusually dominates and requires accurate prediction of the liquid content or hold-up. At highflowrates friction dominates. The acceleration term is usually only of significance at very highflowrates which may be impractical for other reasons such as erosion. The competing hydro-static and frictional losses usually result in a ‘U’ shape to the plot of tubing pressure loss vsflowrate.

Original two-phase pressure drop methods were based on simple correlations but they werefound to be limited when applied to conditions outside the original data range on which theywere developed. New methods based on mechanistic models for the flow regimes provide ahigher level of confidence. The Ansari mechanistic model has been shown to give pressuredrops of within 3% of the Foinaven EWT data. Typical tubing sizes of interest in subsea wellsare in the range 4” to 7”.

(d) Choke and manifold pressure loss

Usually the pressure loss associated with the subsea wellhead, manifold and jumpers etc.. isrelatively low, typically tens of psi compared to the control chokes with hundreds of psi pres-sure drop unless full open for maximum production rate. The pressure drop generated by theproduction control chokes is usually determined by a calibrated correlation for the particulartype of choke and its duty. A representative relationship is usually provided for design andduring operation the performance of the actual correlation used can be monitored through welltesting. It is often the case in steady state analysis that the solution involves calculation from thereservoir to the well head and from the facilities to the manifold, with required choke pressuredrop inferred from the difference in pressures between the well head and the manifold.

(e) Flowlines and risers

The flow regimes in horizontal and slightly inclined flowlines are analogous to the vertical flowpatterns but are complicated by the asymmetry caused by gravity. These have also beendiscussed in Section 1. The pressure drop across the flowline is mainly due to friction effectsand hence usually increases with flowrate, although if the flowline is inclined uphill hydrostaticeffects can be important. Again the mechanistically based models are thought to provide thebest results when applied to real systems and for this purpose we recommend using the BPmechanistic model developed by GRE in 1992. This model reverts to the Beggs and Brill corre-lation under normal slug flow, but has a special model for terrain slug flow and has a modifiedstratified flow and annular flow model.

As the water depth increases the effect of the riser pressure loss becomes more significant andthe shape of the combined flowline and riser characteristic becomes more ‘U’ shaped due tothe increased hydrostatic effects, similar to a well bore. The largest uncertainties are expectedto be in the accuracy of the riser pressure drop prediction because of the potentially large diam-eter (up to 20”) compared with the small well bore sizes on which the models have been vali-dated, and due to the possible complex riser geometry. The larger diameter and inclinationeffects are expected to lead to increased slip between the phases which will increase the liquidhold-up and the hydrostatic pressure loss. This is particularly true for three-phase flows where a

-

Issue 3 - January 1998 7-

Section 20. Deepwater mult@uzseJow BP Multiphase Design Manual

much greater proportion of water is possible than in the homogeneous assumptions used intwo-phase flow models.

Another potentially large inaccuracy can arise from the hysterisis nature of multiphase flow inwhich the flow regime developed in one section of the geometry may persist for sometime indownstream sections. This is particularly true for large slugs and is a major limitation of generalsteady state two-phase design programs in which the topographical segments are treated asindependent.

(f) Arrival pressure

Historically a range of arrival pressures may occur during the lifetime of a development as the oilproduction rate declines, This can be accommodated by delivering to the lower pressure sepa-rators. For example the initial Pompano fluids are produced into the HP separator at 1250 psig,with the initial WHFP being high enough for the flowlines to run at a back pressure of 2000 psigat the platform upstream of a choke giving rise to bubble flow in the flowlines. As the availablepressure subsea reduces the production can be successively switched to the MP separatoroperating at 450 psig and finally the LP separation stage operating at 150 psig. There is hencea large variation in the arrival pressure that can be used to provide for flexibility of operation.

The arrival pressure is usually fixed by the processing requirements although there will inevitablybe some trade-off with the production capacity of the whole system if an integrated approach isused.

20.2.1.2 Network integration

An integrated approach to the system pressure drops is important if an optimised system is toresult from the design. This approach is likely to overcome past failures in the communicationbetween the various disciplines involved. For example, the subsurface analysis is likely toprovide a minimum WHFP in order to give them a design margin, whereas the process designmay adopt a high delivery pressure to provide them with a better margin for compression forexample. The result can be that there is an unrealistically low pressure drop available for theinterconnecting flowline and riser system. In addition, real systems can involve multiple wells,flowlines, and risers where it can be difficult to optimise production without an integrated modelto determine which wells should be choked to what manifold pressure and produced to whichseparator pressure. Figure 6 shows a simple multiple well, single flowline example where thehydraulic characteristics of the wells are different and may be updated in a database using theresults from well tests to adjust for changing operating conditions. It can be imagined that theapproach becomes even more complicated if other operational constraints are involved such aswell bore gas lift. This complication may give rise to the need for computer assistance in makingoperational decisions for optimised performance and for future field development planning, andcan be followed-up further in Reference 2.

By consideration of the various components in the production system it can be seen that thenatural oil flowrate may be maintained or increased by supporting the reservoir pressure,reducing pressure drop from the reservoir to the separator, and by reducing the delivery pres-sure. Maintaining the reservoir pressure can involve gas re-injection, voidage replacement, orwater injection, usually with associated CAPEX and OPEX cost penalties. The pressure lossbetween the reservoir and the tubing can be reduced through improved completion design andstimulation to reduce the pressure draw down across the formation. The pressure loss acrossthe well tubing, flowline and riser may be reduced by increasing the diameter (although too low

-

-



a velocity may lead to hydraulic instabilities causing unmanageable transient flowrates andpossibly high re-start pressures). Tubing and riser pressure losses may be reduced by usingfoams, gas lift or DRA and boosting may be provided by using ESP’s, multiphase pumping, orseabed separation and pumping. Flowline pressure drops may be reduced by using DRA orremoving the water prior to transportation. Reducing the arrival pressure may increasecompression requirements and worsen oil-gas separation efficiency. The optimum solution willinvolve a trade-off between the competing economics.

In some cases pressure boosting or pressure drop reduction may provide economic alterna-tives to maximise production as a solution to the problems faced by natural flow.

20.2.2 Integrity

‘Designing to ensure that the fluids are contained and that the flow is not impeded.’

One of the biggest potential deepwater problems here is related to cold temperature operationand the impact on production chemistry issues such as wax, hydrate, and emulsion formation.Lower ambient temperatures and longer transportation distances are likely to lead to increasedcooling of the fluids, although this may be partly offset by higher wellhead flowing temperatureswhich may give rise to material limitations.

Two modes of operation are possible:

l Normally operate at temperatures above the wax, hydrate, or emulsion formationconditions and provide measures for abnormal operation such as during shutdownsand restarts.

l Operate in the wax, hydrate, or emulsion formation region but use chemicals tosuppress formation or the occurrence of problematic conditions.

The DeepStar project has recognised the importance of assuring flowing conditions and hasestimated that 1 O-30% CAPEX reductions could be realised if chemicals could be developed toavoid flowline blockages. The cost savings are related to moving from expensive bundlepipeline designs to bare pipe options ($ millions per mile saving), removing insulation require-ments ($150,000 per mile saving), plus $ millions per mile if a pigging loop is avoided, with addi-tional OPEX reduction of $1 million for a 36 hr shutdown at 30 mbopd.

Figures 7 and 8 illustrate flowline temperature profiles for a concrete coated pipeline and abundle with a high value of insulation and indicates that high insulation and flowrates arerequired to maintain high temperatures over long distances.

Solutions to the thermal problems revolve around developing inhibiting chemicals, preferablylower cost alternatives to conventional chemicals, and insulation or heating.

The higher pressure operation leads to increased hydrate formation tendency and corrosion.Higher pressure also leads to more expensive pipeline designs and could lead to the use ofHigh Integrity Pressure Protection Systems (HIPPS) if it is not possible or economic to designfor full shut-in wellhead pressures.

Deepwater flowlines are expected to operate at a higher pressure due to the greater pressure

Issue 3 - January 1998-- 9

Section 20. Deepwater multiphase flow BP Multiphase Design Mumal

drop across the riser. To provide acceptable velocities at the riser top within erosion or corro-sion inhibitor shear stripping limits, the velocity in the flowline may be relatively low leading tothe potential for surging, and solids or water deposition, which can have an impact on theflowing capacity, pigging, and corrosion.

At a first glance the integrity issues related to the topsides plant are not expected to be anyworse than conventional shallow water developments because the arrival pressures and steadyvelocities are expected to be similar. Hence, the slug mechanical loads imposed on thepipework and the erosional velocities are expected to be similar. However, there is concernover the potential dynamic effects caused by the large hydrostatic pressure drops possible atincreased water depths, giving rise to the stalling of large slugs as the gas in the flowline packsto provided the increasing hydrostatic pressure. When such slugs are produced into the facili-ties the extra hydrostatic back pressure is no longer needed and is converted to friction as theslug tail is accelerated giving rise to potentially high transient velocities and mechanical loads.Figure 9 illustrates the acceleration of the slug tail. Section 11 of this manual discusses furtherthe estimation of slug forces. Figures IO and 11 illustrate the potential integrity related issues.

A more detailed description of physical chemistry related issues is provided below:

(a) Low temperature limits and effect on physical chemistry issues

Low operating temperatures occur if the fluid is exposed to a cold environment for a significantperiod of time. The greatest effects that deeper water and longer step-out distances will haveare on the lower ambient temperature in deeper water and longer residence time that the fluidwill have in the transportation system, and riser auto-refrigeration effects. If the pressure dropavailable is relatively low such that a large diameter flowline and riser is required, the residencetime will increase for the same production rate, and this will give rise to lower temperatures. It isapparent that for typical systems the operating temperature may be below 90°F even with highlevels of insulation. As rule of thumb 90°F is a minimum arrival temperature for good separationefficiency.

Each of the physical chemistry issues will be introduced below and considered in the context ofdeeper water developments.

(h) Hydrates -

Hydrate formation conditions are typified by low temperatures and high pressures, hence oper-ations in deeper water with longer tie-backs are expected to increase the likelihood of hydrateformation which can reduce the flowing capacity, or in the extreme, complete flowline blockagecan occur which can give rise to safety and containment problems as well as the potential highcost of remediation and production loss.

Hydrates are part of a chemical class of compounds called clathrates and can form at tempera-tures as high as 70°F in hydrocarbon/water systems operating at high pressures. Hydrates aresolids composed of water and gas and are in some ways similar to ice with a frost like or iceappearance. The density of hydrates is similar to ice. However, whereas a pressure increaseover water reduces the tendancy to form ice as the temperature is decreased, a pressureincrease in a hydrocarbon gas/water system increases the tendancy to form hydrates astemperature is decreased. Figure 12 shows a typical hydrate decomposition curve for aoil/water/gas fluid on which is superimposed a pressure and temperature profile for a 50 milesubsea flowline from deepwater wells to a shallow water platform where the ambient tempera-ture increases. It is seen that most of the flowline operates in the hydrate formation region inthis case.

-


Section 20. Deepwater multiphase~ow

The classical approach to avoiding hydrate problems under normal operation is to insulate theflowlines or to use inhibitors such as methanol or MEG. Insulation or heating the flowline isemployed to move the operating line to the right of the hydrate region, whereas chemical inhibi-tion is used to move the hydrate curve to the left of the operating line. Insulation can signifi-cantly add to the CAPEX cost of a flowline ($150,000 per mile) whereas classical inhibition canrequire large quantities (1 O-40% of the water rate) of relatively expensive chemical which canadd significantly to the OPEX cost. Even if the normal operation remains out of hydrate formingconditions, shut-down and subsequent cool-down can cause hydrate formation. If de-pressuri-sation is used to prevent hydrates during extended shut-down periods, chemicals may still berequired during start-up since the flowline pressure often increases faster than the temperature,taking the operating line into the hydrate formation region. In this case there can be a trade-offbetween the methanol storage requirements and the start-up production profile.

By comparison with Figure 8 it is seen that the flowing fluids cool to 70°F in around 30 miles fora production rate of 5 mbd. At higher rates this distance is increased, hence turn-down to lowflowrates needs to be considered and the lowest expected production rate can define thedesign case, depending on the effects of water cut. Figures 7 and 8 indicate that there aresignificant operating regions where the arrival temperature can be below 70°F. However thepressure may also be lower at the facilities, hence the determination of hydrate free conditionsrequires consideration of the flowline temperature and pressure profile, since as illustrated inFigure 12, it is possible for each end of the flowline to be free of hydrates, yet they may formpart way along the length.

The high cost of classical hydrate inhibitors has recently led to a search for lower cost alterna-tives which are required in smaller quantities. The three main types of low dosage additives arekinetic inhibitors (THI), growth modifiers (HGI), and emulsion additives. Each have their ownoperational implications as discussed below:

Kinetic inhibitors

These are added to suppress the formation of hydrates like classical inhibitors but will not workat high sub-cooling (more than 1 O’C) or for extended time periods. This may have implicationsfor three-phase flow wet/gas systems operating at low velocities where the water phase canhave a very long residence time in the transportation system. The active agents in theseinhibitors are polymers which are typically transported by a carrier solvent and are thought to beeffective for gas, oil or condensate fluids.

Growth modifiers

Growth modifiers are active in the water phase and are designed to modify the growth of thehydrate and to encourage the crystals into the hydrocarbon phase. They are effective up toaround 30% water cut and require that the hydrate crystals are transported as solids in thehydrocarbon liquid.

Emulsion additives

These inhibitors work in the oil phase by creating stable water-in-oil emulsions which do nothave a high viscosity and which can be transported as a slurry, hence a hydrocarbon liquidphase is required. These are the newest and least tested low dosage inhibitors.

Each of the hydrate inhibitors have specific implications for processing and these are outlined inTable 1. The required dosage rate of the new inhibitors can be as low as 0.25 w-t% giving treat-



ment costs of around E4/bbl of water compared to MEG and methanol inhibition costs of El 1and f7 respectively. Given the potential cost saving of avoiding insulation and possibly also ofburying subsea flowlines, the new breed of hydrate inhibitors may have a large impact in thefuture since it will be likely that, even with the best insulation, transport over long distances willnot be possible at temperatures above those at which hydrate can form. There is, however, aconcern that the relatively small sub-cooling possible with these inhibitors will be insufficient forlong step-outs and water depths.

The hydrate formation can sometimes be detected by a reduction in the produced waterflowrate as water is being absorbed to form hydrate crystals. In other cases an increased pres-sure drop may be detected. Complete blockage is usually detected by reduced gas flowratesand pressure abnormalities. Some systems are being developed to detect hydrate and waxdeposits on pipewalls and are based on ultrasonic techniques. These may be useful to monitortrouble spots such as uninsulated jumpers for example.

Complete blockages can be difficult to locate precisely and may take sometime to melt usinginhibitors. A preferred method for melting hydrate plugs is to de-pressure the flowline. Howeverthis generally requires that the pressure is reduced on both sides of the blockage simultane-ously as high differential pressures can cause the plug to move rapidly on melting, and cancause structural loads due to impact damage if the flowline changes direction, such as at theriser base for example. To de-pressure a subsea flowline blockage on both sides generallyrequires a flowline loop to the platform, which is a set-up typically used in the GOM as it alsoassists in turndown and allows for flexibility if producing good and bad producers to differentseparator pressures. However, the total subsea production is lost during the de-pressurisation.The Shell Mensa development only has one flowline and depressuring from both ends will befacilitated using a surface vessel to tie-in to the subsea manifold and lower the pressure on thisside.

Melting hydrates using the application of external heat is possible. However care is required iflocalised heating is employed as very large pressures can be generated as the hydrate vapor-ises if it is contained within the plug.

Although we have discussed hydrate formation problems here it should not be forgotten thatmost production operations can suffer from a combination of simultaneous problems which canbe compounded. For example, hydrate formation followed by or concurrent with wax formationcan form particularly persistent blockages which have in the past lead to the abandonment offlowlines.

(cl wax

Problems with wax or paraffin deposition have been identified by DeepStar as the top priorityproblem for deepwater developments. Typical cloud points or wax appearance temperaturesare usually in the range 50 to 100 “F and are not very sensitive to the water cut or the operatingpressure, but are dependent on modelling the operating conditions by using ‘live’, rather than‘dead’ oil samples. Best estimates of the wax formation and deposition characteristics can bedetermined using real fluids in laboratory or field deposition tests.

As flowing temperatures fall below the cloud point small particles of wax may form which arecarried along with the fluid. A temperature gradient is required for wax deposition on thepipewall to occur and this is usually the case with subsea flowlines where the wall temperatureis lower than the fluid temperature. Because of these temperature gradients wax can deposit

12 Issue3Issue3 -- January1998 January1998


--

before the bulk fluid reaches the wax appearance temperature. Further downstream when thefluids have cooled to the temperature of the wall the temperature gradient and wax depositionis reduced, hence a wax deposition profile can exhibit a hump for a long flowline as illustrated inFigure 13. When the fluid and pipe wall has cooled to the same temperature the wax forms bydiffusion in much lower quantities and can be transported by the fluids. One way of tackling thewax problem could hence be to cool the fluids produced fluids to the ambient temperature atthe wellhead.

Potential problems with wax deposition can depend on the amount and type of wax deposited.Soft wax deposits may be carried along with the fluids and may not cause a significantproblem, whereas hard deposits may require mechanical removal using pigging or melting bysolvents or by flushing with high temperature fluids. There will likely be a trade-off between thelost production due to the reduced capacity resulting from paraffin deposition and thefrequency of remediation, which will incur operating costs and deferred production

For example, the Foinaven crude has a wax content of 7.5-8.5 wt%, which begins to deposit atpipewall temperatures below 38°C (100°F). It is expected that 0.52 mm of wax could bedeposited in the flowline and 1 Z-4 mm in the riser during 1 month of operation. The increasedpressure loss is estimated to reduce the flowing capacity to 97%. Hot oil flushing requires atemperature of 10°C above the wax appearance temperature to be effective and it is estimatedthat a flushing period of 12 hours will be required, which if carried-out once a month would givea average capacity of 98.5%.

In practice optimised operation is likely to involve careful monitoring of the wax deposition andthe evaluation of gains from wax control procedures. If operating conditions change rapidly itmay be difficult to determine whether paraffin deposition is responsible for the reduced systemcapacity or an increase in the water cut, for example.

If pigging is used to remove wax deposits care must be taken to ensure that the pig does notbecome stuck. Some operations have benefited from using soft pigs so that some wax is lefton the wall, which acts as an insulator and reduces wax deposition rates, whereas completewax removal could give rise to high deposition rate and high pigging frequencies, not tomention potential wax handling problems at the facilities.

Deeper water operations and longer step-out distances are likely to giver rise to more severeparaffin problems where conventional remediation methods may have problems. For example, itmay be difficult to maintain the temperature of flushing fluids over longer distances and pigginglarge quantities of wax may not be possible due to plugging. For these reasons there is consid-erable interest in the development of wax inhibitors which modify the wax crystals and preventdeposition on the pipewall. These chemicals may not prevent the formation of wax but maymake them transportable.

As with hydrates, the interaction with other solids and inhibitors may be significant in practicaloperations. For example the BP West of Shetlands developments have the potential for paraffindeposition and relatively high sand production rates of up to 10 Ibs per thousand barrels of oil.The effects of the combined wax and sand production are little understood.


Section 20. Deepwater multiphase flow BP Multiphase Design Manual

(d) High viscosity oils and emulsion formation

The viscosity of oils increases rapidly at low operating temperatures, although this may bepartially offset at high operating pressures due to the effect of solution gas which reduces theviscosity. Figure 14 shows typical oil viscosity vs. temperature relationships. Measurements onfield samples at several temperatures can be used to tune physical property correlations inmultiphase design software, and hence to improve the accuracy of predictions. The increasedoil viscosity leads to greater frictional pressure loss which may have implications for long flow-lines.

Some oils are very viscous and can be non-Newtonian giving rise to high pressure drops innormal operation and high re-start pressure requirements, Some of these problems have beenovercome by using chemicals to form stable low viscosity emulsions or by ensuring that a lowerviscosity fluid is in contact with the pipewall with the high viscosity fluid flowing down themiddle, such as in core-annular flow.

The determination of the liquid viscosity is generally more difficult if oil/water emulsions arepresent since the viscosity can be many times higher than the individual phase viscosities. Theviscosity of an emulsion depends to a large extent on the particle size distribution, which is afunction of the flow history and the shear generated during transport through chokes, pumps,and the flowline itself, for example. High shear creates smaller particle sizes and tighter disper-sions which gives rise to increased fluid viscosity as illustrated in Figure 15. Peak viscosities(over 30 times the oil phase value) typically occur near the inversion point where the emulsionreverts from a water-in-oil dispersion to a oil-in-water dispersion. The inversion point usuallyprevails at water cuts in the range 40 to 60 %.

Emulsions generally exhibit non-Newtonian shear thinning flow behaviour so that the viscosity isdependent on the shear rate. Problems can arise following shutdowns due to increased restartyield stresses and the effects of ageing. Longer flowlines with emulsion formation may requireexcessive pressures to restart the flow.

It is hence seen that emulsion viscosities depend on the temperature, water cut, and theflowing history. Typical assumptions of a weighted average viscosity of the oil and water arelikely to grossly underestimate the liquid viscosity if emulsions are present. More realistic modelsfor the viscosity of emulsions relate the ratio of the emulsion/continuous phase viscosity to theconcentration of the dispersed phase modified by some power. This relationship can be tunedwith measured data to improve the prediction accuracy.

A final point on the effect of high fluid viscosities is the impact on multiphase flow regimes andslug characteristics where it has been seen during tests with air and water/glycerol mixtures atSunbury that high viscosities lead to a larger slug flow region. However the slugs are generallysmaller and more frequent.

20.2.3 Delivery

‘Ensuring that the process facilities can handle the fluid delivery from the flowlines and risers.’

The two main issues here are related to the arrival temperature, which if too low can give rise toseparation problems and if too high may require cooling, and to the variations in the flowrates ofthe fluids delivered to the facilities as a result of normal hydrodynamic slugging, terrain inducedslug flow, or resulting from transients caused by changing operating conditions and pigging.



The arrival temperature problems are related to the low temperature issues discussed in theintegrity section which can lead to wax, hydrate or emulsion formation. In addition, good sepa-ration is also assisted by higher temperatures where around 30 “C (90 “F) is a good rule ofthumb and may require the use of heating plant upstream of the separation facilities. A too highan arrival temperature resulting from HT well production can also give rise to problems and mayrequire coolers necessary for compression or dehydration plant.

Some of the most severe potential operating limitations may be imposed by transient flowratesresulting from normal or abnormal operation. The greater flowline length and riser depth leadsto a larger liquid inventory with potentially increased transient surge problems. In addition, theflowline topography and the shape of the riser can lead to severe slugging at relatively highproduction rates. This would make processing the fluids almost impossible and could requirethe use of control techniques such as gas injection, for example, to alleviate the condition.Figure 16 illustrates the delivery related issues.

Further discussion of the effects of deepwater riser configuration on severe slugging and thelimitations of design methods is provided below:

(a) Effect of riser shape on severe slugging

The effect of the deepwater flowline/riser configuration on the hydraulic stability of a system wasinvestigated by performing PLAC transient two-phase flow simulations for various flowline andriser geometries. Three riser configurations were used for the dynamic simulations and areshown in Figure 17. The configurations are all based on a 10,000 ft water depth and considervertical, catenaty, and ‘S’ shapes. The base case flowline configuration consists of a 20 milehorizontal flowline and sensitivities have investigated the addition of a 10,000 ft, lo downhillsloping section to the riser base. All simulations are based on a 12” inside diameter system.

The PLAC simulations are based on a Pompano fluid composition with a GOR of 1200 scf/stb.A constant inlet mass flow rate is assumed, and two cases equivalent to 10 and 20 mbopdhave been modelled using a fixed inlet temperature of 40 “F. The outlet pressure is fixed at1000 psia.

Table 2 shows the results of the simulations and indicates that inclining the flowline down hill loat the base of the riser has a large effect of the stability of the system. For all cases with a hori-zontal flowline the stability limit is between 10 and 20 mbd, but the dip shape of the ‘S’ leads tomuch more severe surges with a horizontal flowline. With the inclined line the surges producedby the ‘S’ riser were not as severe as with the other shapes due to the trapped gas bubblepreventing the riser completely filling with liquid.

(b) Conclusions on the limitations of transient codes to simulate deepwaterflowline/riser hydraulic instabilities

Although general purpose transient two-phase flow simulators offer the most potential forassessing the hydrodynamic stability limits of deepwater production systems there are presentlysome concerns over the limitations of the codes. This is highlighted in Figure 18 which showsflow pattern map for the BHRg lazy ‘S’ riser tests (Reference 3) compared to the results of theOLGA code. It is seen that OLGA predicts a much smaller severe slugging region. Present limi-tations of the transient codes are thought to be:


Section 20. Deepwater multiphase flow BP Multiphase Des&n Manual

Cannot inherently model slug flow

Flowline slugging can have a large impact on the severe slugging region. However generalpurpose transient two-phase flow computer programs such as PLAC and OLGA usually employaveraged equations when slug flow is predicted, so they are insensitive to the flow and pres-sure variations that result from normal hydrodynamic slug flow. OLGA has been modified toincorporate a ‘slug tracking’ model, but this is a simple model that initiates slugs of a pre-deter-mined length when formation conditions occur. In some cases this can ‘force’ the result andwhile it may be a useful tool, it does not necessarily lead to a valid simulation. The implementa-tion of the slug tracking model also slows the code down significantly. Work is in progress onboth PL4C and OLGA to improve the modelling of hydrodynamic slugging.

Poor boundary conditions

Normally when transient codes are used to model severe slugging in flowlines and risers simplefixed inlet and outlet boundary conditions are usually specified. In most cases the inlet massflowrate supplied to the flowline is specified as constant, as is the platform arrival pressureboundary.

In practice the flowrates at the start of the pipeline may vary as a result of slugging or heading inthe wells, or even as a result of flowline back pressure variations affecting the well inflow rate. Ifthere are interactions between the wells and the flowline this may require that the wells aremodelled also, although if the pressure drop across the choke is high this may de-couple thewell and flowline dynamics.

At the other end of the system variations in the arrival pressure may result in practice due to theinteraction with the process plant and its control action. Even relatively small outlet pressurevariations have been shown in the past to completely dominate the transient response of flow-lines. There have been various projects aimed at coupling dynamic process plant simulatorswith transient pipeline codes such as the D-Spice/OLGA interface. This goes some way tosimulate the interaction between the flowline/riser and the facilities, but the resulting tools arerelatively cumbersome and may not be appropriate for simulations early in a project where thereis insufficient definition of the facilities. Under these circumstances a sensitivity to typicalboundary condition variations may be a sensible way to proceed.

Three-phase flow effects

If water is present in the fluids there will be a tendency for slip to take place between the oil andwater phases as well as the gas and liquid phases. This will be more pronounced at lowflowrates typical of severe slugging operating conditions and will give rise to a higher liquidholdup and density compared to the homogeneous liquid phase assumptions usually applied tohandle water cut effects in two-phase flow steady state and transient simulators. In addition,the homogeneous assumption cannot model the actual oil and water arrival rates and this mayaffect the operation of the process plant if, for example, transient sweep-out produces a largewater slug. It is expected that proper accounting of the three phase flow may result in anincreased region of flow instability.

A simple three-phase flow model has been implemented in OLGA and there are plans toinclude one in PL4C also. BP also developed an in-house code TRANFLO which comparedfavourably with dip slugging experiments and may be developed further in the future.

I6 Issue 3 - January 1998

Section 20. Deepwater multiphase flow-

Single composition

Both PlAC and OLGA use a single fluid description for the transient simulation which is used togenerate look-up tables of the physical properties. This is adequate for most cases, but ifmodelling riser base gas injection, for example, the injection gas is forced to have the feed gascomposition. In practice the injection gas will have been processed and will have a differentsolubility in the oil which will effect the free gas content and the overall effect on the systemstability. Compositional tracking has been attempted with both PLAC and OLGA with limitedsuccess due to the much longer computing times involved. The TACITE code developed by IFPhas compositional tracking but is limited to two components at present.

Simple one dimensional flow models employed

The hydraulic models employed in the transient simulators rely on simplified one dimensionaldescriptions of the flow regimes with closure laws derived from correlations. In some cases thisapproach has limitations due to the three dimensional nature of the flow and the inherentcomplexity in real systems, In addition, the closure laws are usually based on laboratory scaledata that may not be valid under field conditions.

Field trials are proposed on Foinaven in 1997/8 that should provide valuable operating data andexperience for the testing of the steady state and transient design tools.

issue 3 - January 1998 17

ISection 20. Deepwater multiphase flow

-


LEFT BLANK


3Section 20. Deepwater multiphase flow

20.3 Deepwater operating limitations and solutions

We shall now combine these operating limits in order to assess those which are a direct resultof the deepwater nature, and those which are more generic, in an attempt to identify the criticalissues for deepwater developments. The most promising solutions are discussed and the impli-cations for future deepwater system design and operation are outlined. It is not possible toconsider limitations simply as low throughput and high throughput limits as the dynamic natureof two-phase flow leads to some obscure results. For example, low steady flowrates may leadto riser slug stalling and accelerations that produce higher loads than steady production at highthroughputs.

A large proportion of the operating limitations are related to high or low flowing velocities whichare partly determined by the characteristics of the reservoir fluid, and are principally beyond ourcontrol, for example the GOR and water cut. However down-hole gas and water separationmay change this in the future. The operating pressure and the flowline size are also major vari-ables.

It can be difficult to assess the relative importance of the various operational constraints in prac-tice as it may not be possible to provide a large enough operating envelope which avoids alllimitations, and it may be required to asses the grey areas of operability which mitigate againstthe worst problems. Such an assessment requires that all the relevant considerations areunderstood and weighed-up. This often requires input from various disciplines in a true inte-grated design.

Figure 19 shows a typical operating envelope that was produced for the Foinaven developmentand considers the capacity limit due to the available pressure drop, the unstable hydraulic oper-ating limit, the erosion rate, sand deposition limit, and the minimum arrival temperature. Thisonly represents a few of the possible operating limitations, and the Figure shows a clear oper-ating envelope. If the water depth and tie-back distance were to increase the flowline pressurewould increase and the arrival temperature would decrease, with the result that some of thelimit lines may move to reduce the operating region, or may cross-over to produce a grey oper-ating region.

The relative importance of the various operating limits depends to a large extent on the specificdevelopment and hence it is not possible to reliably generalise. It is recommended that theresults of this study be applied to a number of potential future deepwater developments inorder that the research efforts can be better focused.

20.3.1 Critical factors for deepwater hydraulic design

The most critical factors for the successful design and operation of Deepwater developmentsare judged to be as follows:

(a) Provision of required system pressure drop

The differential pressure across the transportation system will increase with water depth andtie-back distance, hence requiring higher wellhead flowing pressures and larger flowline diame-ters. The flowline inlet pressure may be increased by:


Section 20. Deepwater multiphase flow BP Multiphase Design Manual

l High reservoir pressurel High well PIl Reduced tubing lossl Low choke pressure drop

The tubing loss may be counteracted by adding energy in the form of down hole ESP’s orreduced by using DRA/foams or gas lift. Gas lift has the disadvantage of potentially increasingthe flowline pressure drop.

The capacity of the flowline/riser system may be increased by:

l Boosting at the wellhead or riser basel Reducing the delivery pressurel Increasing the flowline sizel Using DRA or foamsl Water removal at the wellhead or down-hole

The effect of each mitigating measure should be judged using accurate pressure drop predic-tion methods in an integrated system model. Bearing in mind the practicality and cost of someof the boosting measures, the most promising technologies for deepwater are considered to bedown-hole ESP’s, water removal systems, the use of drag reducing agents, and accurate inte-grated model predictions. The addition of energy using ESP’s is probably the most efficient andreliable means, but may still have problems due to long distance power transmission. Othermeans of supplying power using injection water, for example, may be worthwhile considering.

(b) Low temperature related problems

Low operating temperatures are likely to result from the lower ambient temperature and longerfluid residence time in the transportation system. The low temperature related problems may beovercome by:

l Reduced flowline diameterl Insulationl Use inhibitorsl Chemical or electrical heatingl Piggingl Hot oil and solvent flushingl Water separation at wellheadl Don’t insulate risers

Because of the difficulty and cost of maintaining the temperature of long tie-backs the mostpromising control method is the use of low dosage inhibitors but these may have limited appli-cation due to the relatively low subcoolings that can be tolerated. Cooling at the wellhead maybe a potential for wax problems since if the fluids are cooled to the temperature of the pipe wallthere is no temperature gradient for wax deposition and the small amount of wax that is formedby diffusion can be transported as a slurry. The auto-refrigeration in the deepwater risers waspreviously not anticipated and should be the subject of further study.


Section 20. Deepwater multiphase pow

(c) High temperature problems

High temperature related problems are only likely in the vicinity of the wellheads and are mainlyassociated with increased corrosion and the strength of coatings and flexible pipes. Theseissues are being addressed by HP/HT developments and are not considered as critical here.

(d) High velocity limits

Steady state velocities at the receiving facilities are not likely to be any higher than with conven-tional developments, but the transient velocities caused by slugging could be much higher asthe water depth and the riser diameter increases. This can give rise to intermittent erosion andcorrosion inhibitor film stripping. The biggest problem for deepwater is considered to be in theprediction of slug accelerations and the pipeline load and fatigue implications. If accuratepredictions cannot be made then in-service monitoring of the material wall thickness andfatigue life could be necessary. Loads may be reduced by aerating the slugs or moderating theslug accelerations.

(e) Low velocity related problems

The high back pressure generated by Deepwater risers gives rise to lower operating velocities inthe flowline which may lead to solids deposition and water accumulation. Pigging can be usedto control these problems to some degree, but is costly and time consuming. There is little thatcan be done about the solids problem apart from separation at the wellhead. Remote waterremoval is also a possibility as is the use of emulsion forming agents to maintain the water insuspension in the oil phase, and hence avoid water accumulation and corrosion.

(e) High pressure operation

High pressure operation is likely to be unavoidable. There is still a potential that there will be aconsiderable difference between the wellhead shut-in pressure and the maximum requiredoperating pressure of the flowline. HIPPS systems can be used to reduce the flowline pressurerating and cost if the system is able to monitor the pressures and control effectively. There are anumber of circumstances that can arise where the measurement of the pressure at the ends ofthe flowline is insufficient to infer the pressures throughout, such as during pigging, or if block-ages occur. Here dynamic hydraulic models, tuned to field measurements, may improve theaccuracy of the shutdown system.

(g) Problems with gas and liquid flowrate variations at the facilities

Larger diameter flowlines and risers are likely to generate bigger normal hydrodynamic slugsand the longer flowline length combined with the shape of the riser may increase the likelihoodof severe slugging. The larger system inventory can also lead to more severe surges duringtransient operation. Deepwater systems are hence likely to experience more severe flowratevariations at the facilities than conventional developments. The flowrate variations resulting fromsevere slugging are so large that this mode of operation is to be avoided, so the use of controlmeasures such as riser base gas lift or the use of foams is essential, as are design toolsrequired to predict the occurrence of severe slugging. Methods for limiting the effects of tran-sient operation, such as reducing the liquid inventory and using operating procedures, are also


Section 20. Deepwater multiphase fcow BP Multiphase Design Manual

important. Normal slug flow is more difficult to avoid, and methods for active control and detec-tion subsea may be worthwhile pursuing.

The critical areas discussed above are illustrated on a deepwater schematic in Figure 20.

20.3.2 Assessment of limitations

The real life assessment of critical design limitations is likely to be complicated by the widerange of factors involved, the variation of operating conditions over the life of the field, and thelarge changes in the flowing conditions from the reservoir to the processing facilities. There isalso a compromise to be made regarding the relative importance of various issues which mayreduce one problem while worsening another.

Table 3 shows an assessment of the significance of a number of variables related to deepwaterdevelopments and may be used to rank the configuration of a development with regard topotential hydraulic limitations. The higher the overall score, the more challenging the hydraulicdesign is likely to be. This is an initial attempt at this approach and may be elaborated on in thefuture if considered to be worthwhile.

-

The practical application of hydraulic design limitations can be communicated using designcharts such as illustrated in Figure 19. However the number of limits may be many and suchcharts may become complicated. In addition the operating envelope may change considerablyas the field is depleted and the system characteristics change. For these reasons it is antici-pated that in the future the operating limitations should be computerised and regularly updatedto take account of changing conditions, Because of the many variables it may not be possibleby intuition to determine optimum operations, and intelligent software tools may be required.These could be linked to the monitoring and control system and warn the operator if correctiveaction is required. An example of this could be if the measurements of the fluids into and out ofthe flowline are used to determine the liquid inventory and warn the operator when to send apig to control the build-up of liquid. This type of approach could be important as operationsbecome more complex and more marginal.

22 -~-__ Issue 3 - January 1998


References

1. Fairhurst, C.P,“Hydraulic Design issues for Deepwater Developments”, SPR report numberSPFUTRT/101/97, June 1997.

2 Fairhurst, C.P,“The Virtual Operator-The Concept of an Advanced Monitoring, Control, andOptimisation System for Multiphase Production Operations”, BPX Facilities Engineering,November 1996.

Issue 3 - January 1998~-~ 23

Section 20. Deepwater multiphase flow BP Multiphase Design Munual


LEFT BLANK

2 4 Issue 3 - January 1998

Section 20. Deepwater multiphase pow

Kineticinhibitor

Growthinhibitor

Emulsionadditive

Type ofadditive

Application:

Multiphase

Gas

Condensate

Oil

Dose rate

Yes

Yes

Yes Yes

Unlikely

Yes

Yes

0.25%of water rate

Yes

No

Yes

Yes

2%of water rate

Yes

Yes

0.1250.25%of water rate

1 O-40%of water rate

Operating limits:

Temperature 15°Csubcooling

30-40% WOR

15°Csubcooling

30-40% woFl*

Not known Probablyeffective

Heating / Heating /demulsification demulsification

Demulsifiers Demulsifiers

15°C subcooling(dependant on dose)

10°Csubcooling

No limit No limit

Veryeffective

Water cut

Effectivenessagainst ice

Additionalequipment

Additionalchemicals

Known deploymentproblems

Regenerable

Downstreamimpact

Not typicallyeffective

RecoveryNo

NoNo

Emulsion formationviscosity of blend

No

Not Not Materialcompatibility

80% typicallyUnlikely No

Possible productcontamination

Occasional productcontamination- loss of LNGspecification

Noneknown

Possible productcontamination

potential cause ofhaze in heavier

fractions

potential cause ofhaze in heavier

fractions

In oil

Notknown

Unknown

In all phases

Onshore disposallimits

E

Disposal stream

Environmentalconsideration

UK regulatoryrating for disposal

In water Typically in oil

Water qualityhydrocarbon limit

Notknown

D/E A

* For these additives, the amount of hydrate formed and so the effective operating limit could be a complicated function of flow rate,flow regime, gas amount and water cut. At this stage in development, practical limits are best determined from field trials, or better, fromloop tests

Table 1. Characteristics of hydrate inhibitors

Issue 3 - January 1998 2 5

Riser

Vertical

Vertical

Vertical

Catenary

Catenary

Catenary

‘S’

‘S’

‘S’

‘S’

Flowline

20 mile

20 mile

20 mile with 1 o dip

20 mile

20 mile

20 mile with 1’ dip

None

20 mile

20 mile

20 mile with 1’ dip

Oil rate(mbwd)

IO

20

IO

IO

20

IO

IO

IO

20

IO

Staticanalysis

Unstable

Stable

Unstable

Unstable

Stable

Unstable

Unstable

Unstable

Stable

Unstable

Dynamicanalysis

Small fluctuations - 1 Obbl surge

Stable

Severe slugging - 1069bbl surge

Small fluctuations - 1 Obbl surge

Stable


Stable

Slugging - 306bbl surge

Stable


FVariable Extremes / Effect

c” t Worst t + Best +z0

i: Shallow (Oft)3

Water depth Very deep (10,OOOft)

i7iis Flowline/riser Large 28” Small 3”

2 diameter2

Upstream Long Shortflowline length 100 miles 0 miles

Inclination of O-20” down 7-20” upupstream flowline

Geometry of riser

% of productionup the riser

Double ‘S’ ‘S’ Catenary Vertical

100% small %

Significance(out of IO)

1 o-o

10-l

8 Lowflow - 02 highflow

IO Lowflow - 02 Highflow

IO Lowflow - 2 Lowflow2 Highflow - 2 Highflow

1 o-3

Multiply to get comparisons of different combinations

Om -

1OOOm -

2000m -

3000m -

Gulf of MexicoTroikaNeptuneDiana

m Wind 42.0 m/s20 50

West AfricaGirassol

tm Wind 25.0 m/s ml Wind 38.5 m/s

Hmax 23.2mHs 12.5m

Waves

Surface current 1 .l Om/s

Hmax 7.lmHs 4.0m

Waves

Surface current 1.41 m/s

iMin temp = 4°C

Seabed curent 0.50m/s

Min temp = 4°CSeabed current 0.1 m/s

Water depth 3000m Water depth 2000m Water depth 1500m Water depth 1000m

Northern NorwayNyk HighOrmen Lange

20 50

Atlantic FrontierFoinavenSchiehallionSuilven

-IWind 40.0 m/s20

Hmax 30.0mHs 15.7m

Hmax 32.7mHs 18.0m

Waves Waves

Surface current 1.75m/s Surface current 1.96mk

Min temp = 1.5%Seabed curent 0.63m/s

Min temp = 1.5%Seabed curent 0.49mk

f


\

\ \ i

\, i tl sa

Figure 2. Effect of depth on pressure drop- 12” ID, 1200 GOR, 0% WC, 20 MBD, 1000 psia delivery

Issue 3 - January 1998 2 9

Section 20. Deepwater multiphaseflow BP Multiphase Design Manual

.- ‘iii ‘izTiELna0008888mrnlc7

IIIIIIIIIIIIIIIIIIIIIIIIIIII

: I 1

I i 1I i I I

II

Iii

I@!’!!!!!!!!

:i

I’I8 :

-

Figure 3. Effect of distance on pressure drop- 12” ID, 1200 GOR, 0% WC, 20 MBD, 1000 psia delivery


Limitation

huff icient flowing pressure

Impact of deepwater

l Pressure drop increaseswith water depth andstep-out distance

Solutions-

-

l Accurate prediction modelsl Proper integration modellingl Multiphase pumpingl Gas liftl Separation and boostingl Drag reducersl Foaming agentsl Lower arrival pressurel Subsea water removal


-

Figure 5. Wells as part of the production system


I

Well 1

IWell 1

3 paI 1P----,

e Ii -ri Flowrate Q1

338 I Well 3

kk3

t-

Pl ----1

II

Flowrate Q3

Well 2 Well 3 Well 4

Well 2

I----- 1I

Flowrate Q2

I Well 4

to

III 1I

L

Flowrate Q4

Total flow in flowline = QT = Q, + Q2 + Q3 + Q4

II-----------

//

A i SeparatorI pressure

..-11,1.-.-11,1,1,1,~,~,~,~,~,~,~,

Flowrate QT

Y

Section 20. Deepwater multiphase flow BP Multiphase Design Mun~lal

; ;I- I I 1I:III I i: 1III i i I

Figure 7. Temperature profiles for pipeline with 2” of concrete coating with burial.

-

-

34 Issue3 Issue3 -- Januatv1998 Januatv1998


Figure 8. Temperature profile with low overall heat transfer coefficient of 0.14 btdhr ft* OF.


Pressure(N/m2)

absolute

1,400,000

1,200,000

1 ,ooo,ooo

800,000

600,000

400,000

200,000

0

- n -. - Slug tail velocity

- - - - Pressure behind slug

- Pressure at base

20

18

16

14

12

1 o Velocity0-W

8

6

0 5 1010 15 20 25 30 35 40 45

Time (s)

Limitation

Pipeline blockage dueto cold temperatures

Flow impeded byhigh viscocity

Flow assurance problemsdue to low flowrates.Corrosion due to water puddling.Corrosion under solid deposits.

Limitation

l Colder ambient temperaturesand increasing heat lossaffecting hydrate, wax andemulsion formation.

l Colder temperatures due toauto-refrigeration

l Colder temperatures giverise to emulsion formation

l Flowlines operate at higherpressures and lowervelocities allowing water tobuild up and sand toaccumulate.

l Higher corrosion rates athigher pressure

-

Solutionsl Insulate flowlinesl Use inhibitorsl Add heat

- bundle heating- electrical- chemical

l Maintain high flowratesl Pig pipelines

l Use inhibitorl Keep hot

l Pig flowlinesl Apply minimum

flowrates criterial Inhibit corrosion

Section 20. Deepwater multiphase flow l3P Multiphuse Design Manuul

c.FI

Figure 11. integrity related issues - Avoid impeding flow/contain fluids-


Pressure(psia)

4 0 0 0

13000 -

2000 -

1000 -

H Y D R A T EII From wellhead

Flowline

R E G I O N

I I I I I I

3 0 4 0 5 0 6 0 7 0 8 0 90

Temperature (OF)

Without inhibition or insulation, production fluids in a 50-mile flow line fromdeepwa ters are in the hydrate region for most of the length of the flow line

-0as!c Deposition rate/temperature

.

\e-

0 ‘\I

- - - - - Wax deposition rate

Temperature

Distance form wellhead


Figure 14. Crude oil viscosity as a function of API gravity and temperature.


Section 20. Deepwater multiphzse flow BP Multiphase Design Nanuul

4c

Viscosity ratio = 25

VIS emulsion

VIS clean oil 20

15

0 Data on seal beach crude E-619emulsions from funnel viscositydeterminations

A Data on natural crude oil emulsionswith saybolt viscosimeter

l Data on synthetic crude oilemulsions with saybolt

- n - n Tight emulsion

- - - Medium emulsion

- Loose emulsion

I’ 0’ Ii Im I! I1 I

II

0I I I I I I I

10 20 30 40 50 60 70

% brine in emulsion

Figure 15. Variation in viscosity of crude oil and brine with brine content

42~ .--- ______ Issue 3 - January 1998

-

.-

Limitation

Flowrate variations

Low temperatures

Impact of deepwater

l Slug volumes increased.l Riser geometry can

promote severe slugging,greater pressure andvelocity swings.

l Can lead to plant shutdown

l Cooler temperatures cangive rise to emulsionformation and separationproblems.

l Problems with handlingwax or hydrate increasingcost of treating andcausing downtime

Solutions

l Change topography offlowline/riser

l Use control systeml Gas liftl Foam fluidsl Operating proceduresl Large slug catcherl Compressor controll Process plant dynamic

operation

l Use inhibitionl Insulatel Add heatl Maintain high flowrates

Section 20. Deepwater multiphase flow l3P Multiphase Desi@ Manual

I 1

i IIi I

I I I I I I I I I

Figure 17. Riser shapes used in analysis


Section 20. Deepwater multiphase.flow

0

0

0

0.-

0

++

++

#++

$+ +

4

Figure 18. Lazy S riser, flow pattern map


I*--\ -\\\

\

\\\

.

\.

\Unstable -\

\

\l \

\ �\, Minimum arrival

\ \w‘\ temperature

\l \ �I

Maximum rate

\

Stable

.

\

l \ Erosion rate

Minimum rate (solids transport)7

Gas / Liquid ratio

Critical design elements resulting from deepwater multiphase flow study

ESP’sWater removal

IntegrityLow dose inhibitorsUnderstand riser auto-refrigeration effectsUnderstand dynamic slug loadsMonitor wall thickness and fatigue lifeControl slugsWater removalSand removalOil/water emulsificationHIPPS dynamic model

IAccurate models for severe sluggingAccurate models for transient operationSevere slug prevention using RBGL and foamsDevelopment of operating procedures

l 1998 Foinaven Field trials will provide unique data for improved understanding and model calibration

Section 20. Deepwater multiphase flow BP Multiphase Des@ Manual


LEFT BLANK


RP lMultiphase Des@2 Mamd Section 21. Corrosion in multiphaseflow

Section 21 Corrosion inMultiphase Flow

21.1

21.2

21.3

21.4

21.5

21.6

21.7

21.8

Introduction

Corrosivity

Corrosion Inhibitors

Database

General Flow Effects

ImA.ised Flow Effect

Flow Modelling for Corrosive Design

Research into Corrosion in MultiphaseSlug Flow

References


Section 21. Corrosion in multiphasefEow BP Multiphase Design Manual


LEFT BLANK


Section 21. Corrosion in multiphase$ow

21.1 Introduction

Corrosion in multiphase, carbon steel flowlines is not well understood and the underlying mech-anisms are still under research. Fundamental CO, corrosion research is, at present, restrictedto simple laboratory flow experiments with idealised fluids. In real field multiphase systems theproblem is far greater due to the wide range of uncertainties, i.e. basic flow parameters, gasand crude property variations, effects of flowline geometry and the performance of corrosioninhibitors.

Assessment of corrosion rate and potential is an inexact science and its application requiresconsiderable experience. This section of the Design Manual has been written only to highlightthe awareness of corrosion to the Facilities Engineer. It must be emphasised that reference to aCorrosion Engineer should be made at the concept stage of any flowline or facilities designstudy.

Multiphase flow design schemes and the use of carbon steel flowlines present attractive capitalcost options for field development. The flowline design team must assess these advantagesagainst high operating costs for inhibition, corrosion monitoring and line repair during field life.

A multiphase, carbon steel flowline is always subject to a corrosion risk. In a design, an assess-ment of that risk is required to see whether corrosion may be controlled through a corrosionallowance and inhibition. These specialised tasks combine stress analysis and corrosion engi-neering specialists to examine pipe integrity requirements, corrosion allowance and fluid corro-sivity, Refsl -4 provide more details.

This chapter examines the multiphase flow effects on corrosion and inhibitor performance andwhere they are important in a flowline design.


Section 21. Corrosion in multiphase flow BP Multiphase Design Mammal


LEFT BLANK


i

BP Multiphase LIesign Manual Section 21. Corrosion in multipha.seJow

21.2 Corrosivity

CO, corrosion is caused by a combination of the CO, content in the gas and the presence ofproduced free water. The corrosivity of the fluid is affected by pH, partial pressure of the gas,temperature, H,S content, etc. Simple models are available to assess the corrosivity of a fluidunder flowing conditions, see Ref 5. This is a specialist task for a corrosion engineer as valuejudgements need to be made based on experience.

Issue 3 - January 1998 5-

Section 21. Corrosion in multiphaseflow BP Multiphase Design Manual


LEFT BLANK


BP Multiphase Design Manual Section 21. Corrosion in multiphasefiow

21.3 Corrosion Inhibitors

Corrosion inhibitors are long-chain polymers that reduce the corrosion rate by adsorbing ontothe flowline wall. They are high cost chemicals that fall mainly into either the imidazoline orpolyamine chemical groups. Actual oil field inhibitors contain many additional components,such as sutfactants, in a proprietary formulation. Usually an inhibitor is designed to be soluble ineither the water or oil liquid phase and dispersible in the other phase.

The physical processes governing the distribution of the inhibitor, and its adsorbtion onto themetal surface, are not well understood, especially in multiphase flow. It is known that the timerequired for successful inhibitor adsorption may be counted in minutes, whereas, the timescaleof turbulence activity near the pipe wall surface, e.g. in slug flow, is measured in milliseconds.This timescale mismatch and the intermittent nature of slug flow are considered to be importantfactors affecting the corrosion rate.

The design dosing rates are quoted in parts per million, typically in the range IO-SOppm, whichrequires a known solvent liquid phase flowrate to calculate the inhibitor injection rate. Inevitablythe phase flowrates change so that, unless care is taken, widely differing dosages are deliveredthan those required by the design. In addition, operational problems such as lack of mainte-nance of the injection equipment can severely affect the delivered dosages; episodes where theinhibitor supply tank has run dry are not uncommon. Some success in reducing the corrosionrates to manageable rates, typically 0.05 mmlyr, in inter-field multiphase lines in Prudhoe Bay,has been attained by increasing the dosing rates to 100 - 200ppm.


Section 21. Corrosion in multiphase~ow BP Multiphase Design Ma~zual


LEFT BLANK

8 Issue 3 - Januatyll>

-

Section 21. Corrosion in multiphaseJow

21.4 Database

A database of corrosion rate and flow conditions is available on PC, Ref.6. The system isknown as the Corrosion Rate Assessment dataBase System (CRABS) version2 consisting ofExcel data worksheets containing empirical and erosion rate data and details of the fluid andchemical conditions under which they were measured.

Information on both field and laboratory measurements is contained in the database and maybe used to:

l Establish representative metal loss rates for use in pipeline design

l Investigate for correlation between corrosion and the fluid and chemicalcondition recorded

l Generate data to aid in the validation of computer models

The database is maintained by the Materials and Inspection Group, Sunbury.


Section 21. Corrosion in multiphaseflow

-


LEFT BLANK

-


BP Multiphase Design Munual Section 21. Corrosion in multiphase$ow

21.5 General Flow Effects

Multiphase flow regimes in flowline systems have been discussed in Chapters l&3. If the fluidphases are corrosive, under flowing conditions, the resulting flow regimes can affect the corro-sion rate by varying the corrosion inhibitor efficiency, Ref.7. Generally, stratified flows in horizon-tal flowlines do not present a problem, if adequately inhibited. However, at very low liquid veloc-ities with little or very poor inhibition, considerable damage can occur. Typical damage resem-bles “tramlines“, lines of localised pitting at the liquid gas interface, located at the bottom of theflowline. Sand or solids may remain at the bottom of the flowline and promote under-depositcorrosion. Due to geometrical or terrain changes, the stratified flow may change and alter itscharacteristics radically, this is discussed in the next section.

The slug flow regime, in a horizontal flowline, may have a very large effect on inhibitor petfor-mance. During the passage of the slug, gas is entrained into the highly turbulent slug front and,flow visualisation suggests, is projected downwards to the bottom of the flowline in bursts ofsmall bubbles, Ref.8. High shear stress, the possibility of gas bubble collapse and high turbu-lence are all possible contributors to reducing the performance of the corrosion inhibitor; cur-rent investigations will assist in determining governing mechanisms. Rapid operational tran-sients, e.g. compressor trips or the bringing on of high gas rate wells, may produce very largeslugs.

The annular flow regime has a high velocity gas core which entrains and deposits liquid dropletsfrom and to the surrounding liquid annulus. In a horizontal flowline case the liquid annulus ismuch thicker at bottom of the flowline due to the effects of gravity. There is no evidence thatthe impact of liquid droplets on flowline walls, alone, causes problems. However, combiningsand transport within the high velocity gas core, and a corrosive medium, can produce veryrapid rates of erosion/corrosion, Ref.9.

-


Section 21. Corrosion in multiphase flow BP Multiphase Design Manual


LEFT BLANK


Section 21. Corrosion in mdtiplhaseflow

21.6 hcalised Flow Effects

Flow regimes operating in straight, horizontal flowlines may give rise to corrosion problems inthe absence of other influences. However, it is much more likely that the first signs of damagewill occur at localised flowline features. These would be where details in flow structure andgeometry produce potential corrosion “hot spots“. Likely areas would be at major geometricalchanges, e.g. bends and flowline fittings, tees & wyes. However, minor changes, e.g. weldbeads, gaps at flanges and previous mechanical or corrosion/erosion damage, can also pro-duce serious corrosion problems.

Major geometrical changes to the direction of a flowline increase turbulence production, localwall shear stress and re-distribute the gas and liquid (also solids, if present) phases. The degreeto which this occurs depends upon the acuteness of the direction change, the change in flowarea, especially a constriction, and the proximity of two or more changes. Vertical bends, innominally stratified flow may create very high local turbulence activity as the geometry changeserves to collect the liquid phases, bridging the pipe before being blown through by faster mov-ing gas. Examples of this occurring in a 30“ dia. flowline, in nominally stratified flow, and 40%watercut, created wall thickness losses of 80% and corrosion rates in excess of 500 mpy (12.5mmlyr). Hilly terrain with long upward slopes may produce slugs, which are dissipated in thedownhill sections. Here, again, an analysis of the flowrates without taking pipeline angles intoaccount may predict a stratified flow regime, whereas slugs will have formed somewhere in thesystem.

Minor changes and details to pipewall sections may create local areas of increased inhibitorbreakdown due to the abruptness of the change and water retention. Abrupt wall mismatchesdue to thick walled fittings or field welds of flowline stock manufactured from rolled plate, andthe metallurgical changes due to the heat affected zone, are prime targets for water dropout,inhibitor breakdown and accelerated corrosion. Gaps in flange seals for large diameter flowlinesmay also promote water dropout so that a nominally low watercut of less than 10% could local-ly be subject to, say, 50% or more. Particular care is required if sections of dissimilar materialsare flanged together, e.g. stainless and carbon steels, due to galvanic corrosion, again refer-ence to a corrosion engineering specialist should be sought.


Section 21. Corrosion in multiphase$ow- BP Multiphase Design Munual


LEFT BLANK


BP ,Multiphase Design Manual Section 21. Corrosion in multiphase flow

21.7 Flow Modelling for Corrosion Design

There are no direct methods for predicting enhanced corrosion rates based on the action ofmultiphase flows. The corrosivity of the fluids can be predicted by a corrosion specialist.Inhibitor integrity under flowing conditions, however, is based on field experience and, mainly,laboratory trials, Models, e.g. PROSPER and HILLYFLO, may be used to determine local flowconditions, e.g. velocities, flow regimes and wall shear stresses but the corrosion potential is avalue judgement. For example, in modelling work the shear stresses generated by a slug areestimated by treating it as a piston of liquid travelling at about the gas velocity. This is an over-simplification of the physical processes but allows a degree of quantification and a reference forexperienced corrosion specialists to base a judgement.

The steady-state software described in Chapter 3 may be used to determine the local flow con-ditions and fluid properties, taking into account simple geometrical changes, e.g. slug charac-teristics in long upward and downward slopes. A quantitative assessment of the damagepotential of the slug cannot currently be determined.

Where the geometrical changes are more complicated, such as flow through a tee, or at agreater level of detail, flow in a corrosion pit, the three-dimensional nature of the flow becomesimportant. In this instance, specialist computational fluid dynamics (CFD) software has to beused to predict velocity, pressure, turbulence and temperature distributions. However, again,the corrosion potential is a value judgement.




LEFT BLANK

Issue 3 - January1998


21.8 Research into Corrosion in Multiphase Slug Flow

Research at Ohio University by Dr W.P. Jepson and co-workers has shown that the slug filmFroude number may be used to determine both the severity of the slug and its corrosion poten-tial in the presence of CO,. This technique is still under investigation and is yet to be appliedfully on an operational problem. Froude number analysis of slugs has been based purely onpipelines, i.e. smooth, constant diameter pipes in the absence of localised areas of turbulencecaused by intrusive fittings or junctions.

The most likely effect of an intrusion on the corrosion rate and its possible relationship to theFroude number is shown in the figure below; the shapes of the curves are only approximateand are not based on highly detailed analyses.

rcos

Corrosionrate

rdiff

Possible behaviour I’for an intrusion

Possible behaviourfor the pipeline

Froude Number (Fr)-

Figure1 . Variation of corrosion rate with Froude Number

The rate of corrosion is shown on the y-axis of the trend chart and demonstrates the twoextremes of corrosion. In diffusion controlled corrosion the rate of diffusion controls the rate ofreaction, rdlti and corrosion occurs as CO, attack on the metal surface. However, the reaction islimited by the mass transfer to and from the surface, i.e. the rate of diffusion through the stag-nant and laminar boundary layers is slow, in comparison to the reaction rate.

In reaction controlled corrosion the rate of corrosion is the rate at which the reaction takesplace, rco2. This corrosion occurs as if the metal were placed in a beaker and stirred vigorously,i.e. the turbulence generated by the flow completely dominates the reaction rate.

Issue 3 - January 1998 I 7

Section 21. Corrosion in multiphaseflow BP Multi/,hase Design Manual

-


LEFT BLANK


BP ,Mzdtiphase Design Manual Section 21. Corrosion in multiphase flow

References

1.

2.

3.

4.

5.

6.

7.

a.

9.

“A Corrosion Philosophy for the Transport of Wet Oil and Multiphase Fluids Containing CO,”BP Report ESR.93.ER.013, March 1993

“A Corrosion Philosophy for the Transport of Wet Hydrocarbon Gas Containing CO,”BP Report ESR.94.ER.016, August 1994

“A Summary of Recent Work on Corrosion under Multiphase Flow”BP Report ESR.96.ER.010, January 1996

“The Effects of Low Levels of Hydrogen Sulphide on Carbon Dioxide Corrosion: A review ofIndustry Practice and a Guide to Predicting Corrosion Rates”BP Report ESR.95.ER.073, June 1995

“Influence of Liquid-Flow Velocity on CO, Corrosion: A Semi-Empirical Model”C.de Waard, U.Lotz Paper No1 28, Corrosion 95, Orlando, 1995

“User Guide to CRABS Database and Analysis System (version 2.0)”S.G. Oldfield, RMC Report R97-136(S) November 1997

“Flow Related Damage in Large Diameter Multiphase Flowlines”A.S.Green, B.V.Johnson and H.J.ChoiSPE Journal of Production & Facilities, p97, May 1993

“High Speed Video demonstration - g/9/91 ”VHS video kept by Multiphase Flow Group, SPR, Sunbury

“Erosion Guidelines - Guidelines on Allowable Velocities for Avoiding Erosion and on theAssessment of Erosion Risk in Oil and Gas Production Systems”BP Report ESR.94.ER.070, 1994


Section 2 I. Corrosion in multiphaseflow


LEFT BLANK


BP Multiphase Design Mawal Section 22. Wellhead and pipeline closed-in pressures

Section 22 wellhead and PipelineClosed-h Pressures

22.1 Wellhead Closed-In Pressure

22.1.1 Introduction

22.1.2 WHCIP Calculations for Gas Wells

22.1.3

22.1.4

WHCIP Calculations for Multiphase Wells

The Algorithm to Calculate WHCIP forMultiphase Wells

22.2

22.2.1

Pipe Shut-In Pressure with Upstream Inflow

Introduction

22.2.2 Pressure Increase due to Outlet Closure

22.2.3 Areas for Further Development


Section 22. Wellbead and pipeline closed-in pressures BP Multiphase Design Manual

-


LEFT BLANK


BP Multiphase Des&n Manual Section 22. Wellhead and pipeline closed-in pressures

22.1.1 Introduction

Maximum WHCIP’s have traditionally been calculated using one of the following two models:

1. Assume wellbore contains dry gas only.

WHCIP is then the bottom hole pressure (BHP), which is assumed to rise to the reservoir pres-sure, minus the hydrostatic head of gas. This gives rise to a worse case maximum possibleWHCIP. For wells that produce large fractions of liquid, this model will lead to unduly conser-va-tive predications of maximum WHCIP.

For systems where the tubing may consist predominately of gas when the well is shut-in thismodel must be used.

2. The Classical Equilibrium Model.

This assumes equilibrium conditions between the oil and gas are immediately reached as soonas the well is shut-in. The oil settles to the bottom of the tubing. The pressure at the bottom ofthe oil column then rises to the reservoir pressure. The pressure in the oil column reduces withheight due to hydrostatic head effects. When the pressure in the oil column drops to thebubble point, a sharp interface between oil and gas is assumed to exist. Above the interface adry gas column is assumed to be present. The hydrostatic head loss over the gas columngives the difference in pressure between the bubble point pressure (at the bottom of the gascolumn) and the WHCIP.

Applying this equilibrium model to data from Magnus Platform wells gave rise to an underprediction of WHCIP, i.e. the WHClPs reached during well shut-in exceeded the values calcu-lated using this model. It was believed that the reason why this model under predicts theWHCIP is that on shutting in the well much of the gas present (under relatively low pressureflowing conditions) did not re-dissolve in to the oil. Thus thermodynamic equilibrium did notexist for the oil at the relatively high pressure conditions produced as the bottom hole pressurerises up to the reservoir pressure. Hence, for the shut-in condition, a volume of gas in excessof the equilibrium quantity is left in the tubing. The hydrostatic head of fluid in the tubing is thenless than would be determined using equilibrium assumptions and hence the WHCIP is higherthan predicted using this model.

A new model has been developed to account for this non-equilibrium affect and is discussedbelow.

The two models recommended for determining WHCIP are discussed in detail below:

l The first is for systems where a dry gas column may exist in the tubing on shut-in.(Section 22.1.2)

l The second is for systems containing liquid and accounts for non-equilibriumconditions that can exist in the tubing following shut-in.(Section 22.1.3)

Issue 3 - January 1998- 3

Section 22. Wellbead and pipeline closed-in pressures BP Multiphase Des&12 Manual

22.1.2 WHClP Calculations for Wells that may contain Gas only.

For any system where a column of gas may exist at shut-in this approach must be used topredict the worse case maximum WHCIP. This applies to gas, gas-condensate reservoirs, andalso oil reservoirs which contain a gas cap. Determination of head loss over a gas column iscomplicated by the change in gas density, and hence hydrostatic pressure gradient with height.The simple equation traditionally used to compute hydrostatic head loss over a column is:

Ptop

Mw.g.H= -exp ( 1

Pbottom ZRT

where:

Ptop

= pressure at the top of the gas column (bara)Pbottom = pressure at the bottom of the column (bara)Mw = molecular weight of the gasg = 9.81 ms-*H = height of the gas column (m)Z = average gas compressibilityR = 8314 J K-l Kmol-’T = average temperature (K)

This equation however does not account for the change in compressibility (Z) with pressure(i.e. it requires input of one fixed value of Z). In order to correctly account for the change in Z,and hence in gas density with height and pressure, the following procedure is required:

l Set up a model of the tubing.

l Ensure that the homogeneous correlation will be used for vertical upflow.

l Produce compositional physical properties for the fluids.

Run the simulation at a very low rate, eg 0.1 MMSCFD, so that the frictional pressure drop isminimal (less then 2 psi). The homogenous correlation is employed so that no slippage occursbetween gas and liquid. The simulation should be run with a fixed temperature throughout thetubing. This should be set to the reservoir temperature. The inlet pressure (i.e. the bottomholepressure, BHP) is set equal to the maximum reservoir pressure. By using a simulation tocompute hydrostatic head loss, the change in gas density, and compressibility, with height iscorrectly accounted for. The calculated outlet pressure (WHP) is a first pass estimate ofmaximum WHCIP.

If liquid is predicted to be present in the tubing during this calculation there is a danger that thehydrostatic head loss over the tubing will have been over estimated and hence the WHCIPunder-predicted. In practice, any liquid present will drop to the bottom and may re-enter thereservoir.


BP Mzcltiplnase Design Manual Section 22. Wellhead and pipeline closed-in pressures

To ensure that the predicted WHCIP is conservative, the initial fluid composition used should beflashed at the WHP conditions to remove all liquid that could exist in the well. The remaininglighter gas from this flash should then be used to produce a second compositional datapack.

Re-running the simulation with this new modified compositional datapack will provide a conser-vative estimate of maximum WHCIP.

22.1.3 WHCIP Calculations for Multiphase Wells.

This method is applicable for oil reservoirs where there is no danger of the liquid present in thetubing at shut-in flowing back into the formation to leave a gas column only. This method,known as the MPF method, is discussed below:

The calculation begins by simulating the well under maximum flow rate conditions and thehighest temperature gradient, to determine the maximum gas content of the well. The pressuredrop method used should be the one that most closely matches actual data for the field inquestion. If no data is available, then a pressure drop method likely to overpredict the APshould be used to ensure that a low wellhead flowing pressure is predicted, so that the gascontent of the well is overpredicted. The flowing conditions used in this initial simulation shouldhave a zero water cut in order to maximise gas content. The mass of gas present in the wellbore under these maximum flowrate conditions is calculated. This data is not directly availablein most flow simulation packages. The mass of gas present can be calculated by consideringthe liquid hold-up in each section and taking an average gas density for that segment. In orderto simplify the calculation a spreadsheet model has been developed. This model is availablefrom SPR Transportation. The model is based around a non-equilibrium concept where thegas in contact with liquid at steady state flowing conditions does not have time to equalise withthe oil falling back down the well. This leads to the fundamental assumption of the model - thatwhen the well is shut-in the mass of gas present remains the same as that calculated underflowing conditions. This is a conservative assumption as in practice some of the gas will re-dissolve in the oil as the pressure rises. The bottom hole pressure is assumed to rise to thereservoir pressure. The mass of gas settles at the top of the well bore. Below this gas columnis the oil column.

A WHCIP must be guessed and then the pressure due to the hydrostatic head of the gas canbe calculated. (Integration is used to eliminate undue errors ) This is the pressure at the inter-face between the oil and gas.

This pressure now becomes the starting point to calculate the bottom hole pressure at thebottom of the oil column. This is done by assuming a constant oil density, determined by thereservoir conditions. A check must then be made between the calculated bottom hole pressureand the reservoir pressure. If they don’t match then the chosen WHCIP is wrong and must berecalculated.

A simplified flow diagram is shown overleaf.

Issue 3 - Jarwan/ 1998 5

Section 22. Wellhead and pipeline closed-in pressures l3P Multiphase Design Manual

22.1.4 The Algorithm to Calculate WHCIP for Multiphase Wells

Steady state simulation ofwell at maximum flowrate.

No water present.

L

I Increase iterations counteri = i + l

Calculate height of gas column:Mi = Ah.A.p, (1)Pi = Pi., + Ah.g.p, (2)No, = No, + M, (3)hi = hi + Ah (4) I’ 1

Guessed WHCIP (PWHcIP)

p, = PWHCIPMm, = 0

Find mass of gas present (Mgas)M sas = sum of gas contained in all

sections of the well.

h, =0

ASet iteration counter to 1, (i=l)

I I

Is M, 2 Mgas ?

Calculate height of liquid andbottom hole closed in pressure

H,,, = Well height - hPWHCIP = Pi + “Ii, 9 Pliq

I 1s (PWtiClP - P,,,) c ToleranceI

INo -,is the actual well head closed-in

-


BP Multipbuse Des@z Manual Section 22. Wellhead and pipeline closed-in pressures

22.2 Pipe Shut-In Pressure with Upstream Inflow

22.2.1 Introduction

When flow lines are shut in at the downstream end it is not always possible or desirable to stopproduction into the upstream end of the line. For example line packing in gas export systems,and lines with downstream pressure relief systems.

It is therefore often necessary to simply shut off the exit to the pipe for some operationalreason, such as a plant trip. The pipeline pressure is therefore allowed to rise as more fluidenters the pipe.

This can be simply represented by treating the pipeline as a tank in which liquid accumulatesand reduces the volume. This results in an increase of the pressure in the pipeline, as well as anincrease in liquid volume.

Below is outlined a method in which this localised increase in pressure can be calculated.

Consideration must be given to the resulting change back to flowing equilibrium conditionswhen the pipe is restarted as transient events will occur, often resulting in larger than normalliquid slugs.

It is important to note here that to fully understand the phase distribution that can exist in a pipewhen it is shut-in, a full transient code such as PLAC or OLGA should be used. These codesalso have the advantage of fully modelling thermal and hydrodynamic effects as well as of beingable to simulate the restart of a pipeline. However, these codes require a large quantity of inputdata and take a considerable time to run.

22.2.2 Pressure Increase due to Outlet Closure

A spreadsheet program based in Excel has been developed to help in the above calculation asan alternative to full transient modelling.

This spreadsheet is called Pipeshut and is available through SPR Transportation.

Pipeshut requires the following information:-

1. Pipeline dimensions (ID and length).

2. Densities of the gas and liquid phases under flowing conditions.

3. The flowrates of the gas, oil and water.

4. The initial temperature, pressure and holdup of the pipeline under normal flowingconditions.

5. Correlations for calculation of Rs, 60 and Ov.


Section 22. Wellhead and pipeline closed-in pressures BP lUultipbnse LIesigll Mmuuil

The following assumptions are made in the calculation:

1. Pipeshut does not model any heat transfer and assumes that the pipeline remains at aconstant temperature.

2. The flowrates of oil, water and gas into the pipe are assumed constant and thereforeunaffected by the increase in pressure. The amount of gas dissolved in oil is howeverconsidered.

3. Ideal gas is assumed throughout.

The calculation procedure is as follows:

1. Calculate the mass of gas and liquid in the pipeline from the flowing conditions andholdup, at the time of initial shut-in.

2. Calculate the velocities and densities of the flowing gas and liquid phases, using theRs, Bo and Ov correlations.

3. Calculate new mass of gas and liquid in the pipeline after a time step, and thenew holdup.

4. Calculate the pressure in the gas phase after a time step using the ideal gasequation, with the gas volume, molecular weight, temperature and mass of gas.

5. Repeat from step 2, for as many time steps as required.

22.2.3 Areas for Further Development

Areas where the code could be made more sophisticated and accurate are:

l Calculating gas compressibility at each calculation step.

l Allowing a mass flowrate which varies with pressure (inlet pressure).This would follow a well performance curve.

l Include thermal effects.

l Adding the ability to model fluids compositionally.

l Providing a starting point for the depressuring of a pipeline. (Same assumptions aremade for this program as those made in the depressuring program.)


Section 23. Pressure relief systems involving multiphase~ow

Section 23 Pressure relief systemsinvolving multiphase flow

23.1

23.2

23.3

23.4

23.5

23.6

23.7

23.8

Introduction

Defining the reasons for requiring arelief system

Describing the elements of the completerelief system

Design for the ‘worst case’ relief scenario

Ensure integration of all design aspects

Assess sensitivity to uncertainty

Check critical aspects of design

Final specifxation of reliefcontrol mechanism, downstream pipeworkand disposal mechanism


Section 23. Pressure relief Systems involving multiphase~ow BP Multiphase Design Manual


LEFT BLANK


Section 23. Pressure relief systems involving multiphase flow

23.1 Introduction

The intent of this document is to concentrate on the specific issues arising from needing torelieve pressure from multiphase lines of significant length, and also to cover issues arising fromhaving multiphase flow in any relief/disposal system, from flowlines or from process plant.

For a full treatment of the design and operation of relief systems please refer to the ReliefSystems Team, Engineering and Operations Support, Oil Technology Centre, BP Oil, Sunbury.Everything contained within this document should be applied &in the context of, and as asupplement to, a comprehensive relief system analysis performed by suitably qualified individ-uals or organisations.

This section should also be read in conjunction with Section 8 De-pressuring and Blowdown, inwhich particular aspects relating to the generation of cold temperatures due to gas expansionare given a fuller treatment.


Section 23. Pressure relief Systems involving multiphase flow RP Multiphase Design M~mual

.-


LEFT BLANK


Section 23. Pressure relief systems involving multiphase$ow

23.2 Defining the reasons for requiring a relief system

The principal reason for requiring a relief system is to prevent overpressure of equipment and tokeep the equipment within its design envelope. Additionally, blowdown may be a requirementto protect against external fires. Any release of inventory from the system that results fromoverpressure relief or from blowdown must be discharged in a safe manner.

Preventing over-pressuring

In oil and gas production over-pressuring could arise when the outlet from a system or line isshut in (e.g. by valve closure due to process upset), yet the supply continues to flow into thesystem due to failure, or absence, of the supply shut-in control system. Integrity could becompromised, with possible loss to the environment, if the maximum pressure seen by thesystem at any point exceeds the maximum allowable operating pressure (MAOP) of the systemor any component thereof.

The time taken for the system pressure to build up to the MAOP is an important variable. Therelief rate to prevent further pressure build up must be specified for the different possible failurescenarios. The relief rate is likely to be steady (on average over several minutes) until the sourceof supply can be shut-off. A relief control system is required to activate at a suitable pressure,and pass a flowrate sufficient to protect the overall system, such that MAOP is not exceeded.

In some cases it may be possible to protect pipes or vessels from the possibility of over-pres-sure by using instrumentation of acceptable integrity. This is designed to reduce to acceptablylow levels the risk (multiple of frequency of occurrence and consequence) of a high pressuresource being opened to a closed system which cannot contain the full shut-in pressure. Insuch a case a relief system may not be required.

Note: HIPS, HIPPS, Cat 1, 2a . . . are now old terminology. BPX follows the risk based methodannotated in the UKOOA guidelines referring to SlLs (safety integrating levels)

The judgment of whether to invest in full pressure containment, lower pressure rating withinstrumentation of acceptable integrity, or lower pressure rating with relief, must be made withfull knowledge of the risks and costs associated with each option.

Inventory removal

In case of fire or other emergency that may lead to over-pressure a relief valve may be desir-able, but only if a non-reclosing device is used. This is rare. We would normally use a blow-down valve separate to the PSV or process equipment through a controlled disposal mecha-nism, rather than run the risk of large hydrocarbon inventories in the affected vessel/area beingreleased if there was a leak or rupture due to the over-pressure.

In such cases the blowdown rate is likely to have an initial peak with subsequent decrease tozero as the system, which has no supply after shutdown, de-pressurises to a lower pressuresetting or to atmospheric pressure.


Section 23. Pressure relief Systems involving multiphasepow RP Multiphase Design Manual

Establishing possible relief scenarios for the complete system

It is necessary to determine the cases under which the relief system could be called to operate.As stated in the introduction, this is specialist work requiring expert help (quantified risk assess-ment). Contact the BP Oil experts directly, or via the Multiphase Group on request.

However, by way of introduction, these are the basic questions that need to be addressed.

What is the range of flowrates from the source?

The range needs to be specified as a function of time from start of relief, and, in the case of oiland gas wells, also as a function of time from commissioning (to take into account the likelychanges in reservoir performance through the years of production). The flowrate range willaffect the number and rating of relief devices in the system.

Is the disposal system connected to more than one source?

If so then both the frequency of each source being connected to the relief system, and the like-lihood of multiple sources being connected to the disposal system at the same time, needs tobe determined. This is vital in the determination of the maximum required capacity of thedisposal system.

The answers to these questions will begin to form the design case for the relief and disposalsystems. Detailed design will, however, need substantially more detailed investigation.

-


Section 23. Pressure relief systems involving multiphase flow

23.3 Describing the elements of the complete relief system

In order to design a relief system, it is necessary to predict correctly all the interdependenteffects in order to ensure safe operation, and to have sufficient detail of all the elementscomprising the system.

For any system it is necessary to specify the fluids properties over the full range of pressuresand temperatures that could be seen during a relief event. It is possible to make simplifyingassumptions, but the latest models for relief valve performance are dependent on the fluidproperties so the more accurate the fluid compositional representation, the more reliable will bethe prediction of system performance.

Any one part of the system will be affected by the way the fluids behave in another part.Slugging flow in a multiphase line being relieved will arrive at the relief valve, which then will seewidely varying liquid and gas flow rates. The relief valve imposes a back pressure on thesystem being relieved, thereby affecting the flowrates delivered. The route from relief valve todisposal may also impose a significant back pressure on the relief valve which affects the fluidsflow rate it can pass. The generation of slugs in the downstream pipework will require design ofadequate pipe supports, appropriate design of the liquid knock-out system to achieve therequired performance and careful design of the disposal system (e.g. to prevent liquid carryoverinto the flare).

It is therefore very important to carry at out an integrated assessment of the system over thewhole range of relief scenarios to ensure that operation of this ‘safety’ system does not itselfresult in damage or loss.

Interaction with any relief valve vendor is important, to ensure they have full understanding ofthe conditions under which their product may be called for relief device selection to operate.


Section 23. Pressure relief Systems involving multiphase flow l3P Multiphase Design Manual-


LEFT BLANK


-

Section 23. Pressure relief systems involving multiphase$ow

23.4 Design for the ‘worst case’ relief scenario

What is the worst case?

In many systems the ‘worst case’, in terms of behaviour of the relief system, and its ability toprotect the upstream elements without damage to itself or overloading of the disposal mecha-nism, is the case of the highest flowrates requiring relief.

There are also instances (such as the upstream pipework being long multiphase lines over hillyterrain) when low flowrate cases can be more severe in terms of ensuring protection againstover pressurisation of the upstream system. Lower flowrates may also be more problematic interms of behaviour of any multiphase flow occurring in the pipework between the relief devicesand the disposal system. This can be true even when the flow arriving at the relief device issingle phase.

So it is necessary to determine the worst case scenario for the system. As previously stated,from a process plant upset and shutdown, requiring inventory reduction or removal, then thehighest flowrates experienced by the disposal system are likely to be just after the shutdown.Rates then decrease with time as the inventory reduces. Depending on the shutdownsequence, high rates may persist for some time, as a rate decrease from one part of theprocess is balanced by new flow from another part.

The natures of flow arriving at the relief and disposal systems need to be determined for arange of relief rates. If slug flow arrives at a relief valve, for example, then the pressure dropacross the valve, and the flowrates the valve will pass, will vary with time. The length and liquidcontent of the slug may then determine the ‘worst’ case, which might not be at the highestflowrate.

A further extension to this occurs in the specific example of relief from hilly terrain flow. In longhilly terrain lines a plot of pressure loss across the line vs. flowrate through the line has aminimum. At flowrates less than the minimum pressure drop, the pressure loss is higher due tothe dominance of gravitational over frictional pressure loss. So in such an instance when therelief system is required to protect a line, the highest pressures seen at the start of the line mayoccur at low flowrates. The relief system must lift at a condition such that the start of the line isprotected from exceeding MAOP over all flowrates.

A logical step is to protect the system by installing relief at the most likely point of maximumpressure, or by installing a suitably reliable instrumented system to ‘eliminate’ the over-pressurerisk. However, where reliability of an instrumented system and the access, logistics, and practi-cality of installing a disposal system are in question, then the compromise of having down-stream relief may be the most satisfactory option.

Addressing the critical aspects of design

The issues which must be addressed having a requirement for detailed multiphase under-standing and analysis are listed below.

l The flow rate vs. time that must be relieved to protect each element of the system.

l The condition at which the relief system must come into operation.


Section 23. Pressure relief Systems involving multiphase flow BP Multiphase Desigll Mumd

l The back pressure exerted on the relief control systems by the downstream pipeworkand disposal mechanism for the various relief scenarios, and as a function of time.

l The nature of the flow arriving at the relief control system (steady, transient, cyclic . ...).

l The type, size and control method for the relief control system to allow the requiredflowrates to pass so the system is protected, taking into account the pressureupstream of the relief control system, the fluid properties (which will be time varying),the nature of flow arriving from the upstream pipework, and the back pressure exertedon the relief control system by the downstream pipework and disposal mechanism.

l The potential for erosion damage.

l The characterisation of any slugging in upstream and particularly downstreampipework to ensure adequate design of pipe supports.

l The changes in fluid temperature particularly across relief device, to ensureappropriate metallurgy is used.

l The maximum predicted gas and liquid flowrates, to ensure that these areaccommodated by the disposal mechanism.

To address the critical aspects of the system design listed above, it is currently necessary touse a variety of methods and software. For the relief rates from process plant SimSci’s VISUALFLARE or Hypertech’s FLARENET could be used. For pressure drop along lines to the reliefdevice, and then on to the disposal system, INPLANT, PIPESIM, PIPEPHASE or any othersteady state multiphase simulator could be used. One issue here though is the prediction ofthe onset of critical flow in the low pressure line to the disposal system. Some programs maynot use appropriate, or even any, checks on critical flow, and it should be noted that criticalflow for two phase mixtures can occur at a much lower flowing velocity than for either of thesingle phase components.

The nature of flow in the lines upstream and downstream of the relief device may need to besimulated using both the steady state multiphase programs previously mentioned, and the tran-sient simulators PLAC and OLGA. Slug sizes due to normal, slug flow terrain effects, or tran-sients, will need to be ascertained.

Across a relief valve, the fluid behaviour in terms of pressure loss can be modelled using one ofa number of methods (homogeneous, DIERS, . ..). For steady multiphase flow arriving at a valvethe currently preferred method is DIERS for which a sizing Excel spreadsheet is available fromthe BP Oil Relief Team. However, the behaviour of a relief valve with slug flow arriving at its inlethas not undergone any experimental observation, and only some very preliminary modelling hasbeen undertaken using PtAC.

Incorrect sizing of the valve in either possible direction can cause severe difficulties. If under-sized then the pressure may not be sufficiently relieved, and over-pressure may not beprevented. If oversized then slugging issues may arise as the valve opens and closes, and theflow starts and stops. Such opening and closing is also not good for long term valve reliability.

Downstream of the relief device the nature of the flow is generally assumed to be given bysteady state multiphase programs. However, the behaviour of a slug arriving at a relief valve, -

10 Issue 3 - Januaty 1998

Section 23. Pressure relief systems involving multiphase~ow

then passing through and experiencing a very large pressure drop, is not well defined. It islikely that large amounts of dissolved gas would flash off in, or just downstream of, the valve.Hopefully this will cause break up of the slug. The question of whether a slug can re-form in thedownstream pipework is a very important one. If slugs do form, or re-form they could be travel-ling at very high velocity. If the design case is a high flowrate relief then it is likely that theflowing velocities could be too high to sustain slug flow.

The performance of any downstream knock-out pot can be assessed (assuming no slugging)using PROSEPARATOR software. Flare backpressures will need to be obtained as a functionof flowrate from the manufacturer (assuming no slugging).

Issue 3 - January 1998 I I

Section 23. Pressure relief Systems involving multiphase flow BP Multiphase Design Munual


LEFT BLANK

-


Section 23. Pressure relief $ystems involving multiphase flow

23.5 Ensure integration of all design aspects

Assessment of the different elements of the design will be undertaken using a variety ofmethods and software, as described above. These various computer programs are not yet welllinked from a software point of view. There is no direct feedbacWfeedforward between theprograms, so a once off calculation is not possible. Several iterations through the differentprograms may be required to ensure the overall ability of the system to relieve the requiredflowrates and give adequate protection.

As yet there is no software package which can incorporate all the required models for differentparts of the system under different relief scenarios. It is therefore important to integrate all infor-mation obtained and generated to ensure the relevant interactions (especially due to backpres-sures) are thoroughly understood and taken into account.

If a source flowrate is identified, then the different software packages will initially be used todetermine pressure loss in upstream pipework, pressure loss across a given type and size ofrelief valve, and pressure loss in downstream pipework and disposal system. If the systemdoes not generate a consistent overall pressure loss from supply to discharge then iterations onrelief valve size or type, and upstream/downstream pipe diameters, CAPEX of pipe and vesselsizes must be undertaken until a satisfactory solution is obtained.

The extent of effort required here will depend on whether the work is retrofitting into an existingsystem, or a completely new design with fewer existing constraints.

It is very important that an individual is nominated to carry the principal coordinating role. Allinteractions between sections of the system need to be covered, to ensure that the action orresponse of one part of the system is not detrimental to the action or response of another part.This integrating role should also be the one to work as much as possible towards designing outthe requirement for multiphase relief.


Section 23. Pressure relief Systems involving multiphase$ow BP Multipbnse Design AJanuul


LEFT BLANK


Section 23. Pressure relifzf systems involving multiphase flow

23.6 Assess sensitivity to uncertainty

All of the preceding argument assumes that input data, especially in terms of flowrates and fluidproperties, is fixed and accurate. In reality these variables may have a degree of uncertainty forwhich the significance on the design of the over-pressure protection system needs to beassessed. We have to ask by how much does the predicted behaviour of the relief system, andits ability to protect, change with variations in input data which have ranges of uncertainty.What is the worst realistic combination of the uncertainties?

Variations in possible flowrates of gas, oil and water from the source as a function of monthsand years should be checked for the influence on total mass flowrate and slugging tendency.

Water and gas fraction variations will also affect the performance of any relief valve (because ofchanges in fluid properties), which should be checked across the expected range of likelyvalues.

Best judgment should be made of what is the worst realistic combination of the uncertainties, inorder to develop the worst case.-


Section 23. Pressure relief Systems involving multiphase flow BP Multiphase Design hfunual


LEFT BLANK


Section 23. Pressure relief systems involving multiphase pow

23.7 Check critical aspects of design

If the ‘worst case’ can still be accommodated in the design basis generated then the design issufficient. If the ‘worst case’ is not compatible with the design then initial work should focus ona better definition of the input data. If this is not possible then the design should be modified tosuit, with note taken to ensure any CAPEX increase is not disproportionate to the cost ofgetting better input data.

In terms of improvement of design methods, the behaviour of relief valves under multiphaseflow, and slug flow in particular, requires investigation, whether through DIERS or a separatejoint industry research project. The integrity of the relief system will often depend on slugs notbeing present in the downstream pipework to the disposal process, especially if slugs arearriving at the inlet of the relief device, so tracking of slug behaviour through the system is alsoimportant.

issue 3 - January 1998 I 7

Section 23. Pressure relief Systems involving multiphase flow BP Mzdti/hise LIesign rkf~l~lzcal


LEFT BLANK


Section 23. Pressure relief gxtems involving multiphase~ow

23.8 Find specification of relief control mechanism,downstream pipework and disposal mechanism.

Once all issues listed above have been addressed, and this information linked into the specialiststudies of relief system experts, then final specification of the relief system can be made.

Much of this process should be repeated when additional flow/pressure sources are to beconnected into the system.


Section 23. Pressure relief Systems involving multiphase flow BP Multipbase Design Manrurl


LEFT BLANK

-

-

2 0 Issue 3 - January 1998

DEEPSTAR MULTIPHASE DESIGN GUIDEDISCLAIMER

THE MULTIPHASE DESIGN GUIDE INFORMATION IS PROVIDED BY BPHEREUNDER ON AN ‘AS IS’ BASIS. BP AND ITS AFFILIATES MAKE NOWARRANTIES, EXPRESS OR IMPLIED, WITH RESPECT TO THEMULTIPHASE DESIGN GUIDE INFORMATION, INCLUDING WITHOUTLIMITATION, ANY WARRANTIES ON MERCHANTABILITY OR FITNESSFOR A PARTICULAR USE, AND ALL SUCH OTHER WARRANTIES AREHEREBY EXCLUDED.

BP AND ITS AFFILIATES SHALL NOT BE LIABLE IN ANY WAY TO ANYTHIRD PARTY, INCLUDING, WITHOUT LIMITATION, THE DEEPSTARPARTICIPANTS, FOR ANY LOSS, DAMAGE (DIRECT, INDIRECT,CONSEQUENTIAL OR INCIDENTAL), PERSONAL INJURY, ILLNESS ORDEATH ARISING DIRECTLY OR INDIRECTLY OUT OF THE USE OF, THERESULTS OF USE OR THE INABILITY TO USE THE MULTIPHASE DESIGNGUIDE INFORMATION IRRESPECTIVE OF WHETHER SUCH LIABILITYARISES PURSUANT TO AN AGREEMENT, IN TORT OR STATUTORY DUTY.

BP AND ITS AFFILIATES SHALL NOT BE LIABLE IN ANY WAY TO ANYTHIRD PARTY, INCLUDING, WITHOUT LIMITATION, THE DEEPSTARPARTICIPANTS, FOR ANY CLAIMS, DEMANDS, LIABILITIES, ACTIONS,SUITS OR PROCEEDINGS OF PATENT OR COPYRIGHT INFRINGEMENTARISING OUT OF THE MULTIPHASE DESIGN GUIDE INFORMATIONIRRESPECTIVE OF WHETHER PURSUANT TO AN AGREEMENT, IN TORTOR STATUTORY DUTY.

THIS INFORMATION IS PROVIDED WITHOUT PREJUDICE TO THE RIGHTSAND OBLIGATIONS UNDER THE DEEPSTAR JOINT INDUSTRYPARTICIPATION AGREEMENT, ESPECIALLY THE PROVISIONS RELATINGTO THE CONDITIONS OF CONFIDENTIALITY.

bp - multiphase design manual

Documents